WolfBench

Wolfram Ravenwolf’s Five-Metric Framework · based on Terminal-Bench 2.0

One score is not enough.
Because performance is a distribution, not a point.

Most benchmarks report a single average. WolfBench shows five metrics that tell the full story – from the rock-solid base of tasks solved every time, through the average, up to the ceiling of everything ever solved – plus the best and worst single runs that frame the spread. Together, they reveal what no single number can: how consistent an AI agent truly is.
Learn more ↓

★ Ceiling (ever solved)▲ Best-of (peak run)∅ Average (mean score)▼ Worst-of (lowest run)■ Solid (always solved)

📸

👁

GPT-5.6 SolGPT-5.6 TerraGPT-5.6 LunaGPT-5.5Claude Opus 4.6Claude Opus 4.7Gemini 3.5 FlashGLM-5.2 [W&B Dynamo]GLM-5.2 [Z.ai]GPT-5.4GLM-5.2 [W&B]Claude Fable 5 – 13 refusalsClaude Sonnet 5Claude Sonnet 4.6Qwen3.8-27B [W&B]Kimi K2.6 [Moonshot AI]Kimi K2.7 Code [W&B]DeepSeek-V4-Pro [W&B]Kimi K2.6 [W&B]Gemini 3.1 Pro PreviewMiniMax M2.7NVIDIA-Nemotron-3-Ultra-550B-A55B [W&B]Kimi K2.5 (int4) [W&B]Kimi K2.5 (nvfp4) [W&B]GLM-5-TurboGemini 3 Flash PreviewGLM-5.1 [W&B]GPT-5.3-CodexNVIDIA-Nemotron-3-Super-120B-A12B-FP8 [W&B]Gemma 4 31B [W&B]GPT‑5.4 miniGemini 3.1 Flash Lite PreviewGPT‑5.4 nanoMistral Small 4 119B A6Bgpt-6-astraDeepSeek-V4-Flash [W&B]

⇅

Terminus-2Claude CodeHermes AgentOpenClawCursorCodex

0%10%20%30%40%50%60%70%80%90%100%

GPT-5.6 Sol

WI 84.6
🧠 max

■71% ▼83% ∅85% ▲88% ★97%

WI 82.9
v2026.7.7.2
🧠 max

■69% ▼78% ∅84% ▲89% ★93%

WI 86.3
0.144.4
🧠 max

■78% ▼83% ∅87% ▲89% ★94%

GPT-5.6 Terra

WI 81.0
🧠 max

■65% ▼79% ∅82% ▲87% ★93%

WI 81.0
v2026.7.7.2
🧠 max

■67% ▼80% ∅81% ▲84% ★94%

GPT-5.6 Luna

WI 75.7
🧠 max

■57% ▼74% ∅77% ▲79% ★91%

WI 77.2
v2026.7.7.2
🧠 max

■56% ▼75% ∅79% ▲84% ★93%

GPT-5.5

WI 77.1k=3
🧠 xhigh

■64% ▼75% ∅77% ▲79% ★91%

WI 68.8k=4
🧠 medium

■49% ▼67% ∅69% ▲72% ★87%

WI 73.8
v2026.4.23
🧠 xhigh

■61% ▼72% ∅74% ▲78% ★85%

WI 71.8
v2026.4.23
🧠 medium

■55% ▼71% ∅73% ▲75% ★85%

WI 68.5k=4
2026.4.23
🧠 off

■52% ▼65% ∅69% ▲72% ★83%

WI 77.7k=3
2026.04.17
🧠 high

■66% ▼74% ∅78% ▲80% ★89%

WI 77.5
0.125.0
🧠 high

■61% ▼78% ∅79% ▲80% ★91%

Claude Opus 4.6

WI N/A
🧠 off

■63% ▼69% ∅72% ▲75% ★81%

WI 61.1k=3
🧠 max

■49% ▼61% ∅61% ▲62% ★73%

WI 63.4k=3
2.1.63
🧠 high

■51% ▼60% ∅63% ▲67% ★76%

WI 64.7k=3
v2026.3.30
🧠 off

■52% ▼62% ∅64% ▲67% ★79%

WI 59.0k=2
2026.3.11
🧠 max

■53% ▼58% ∅59% ▲60% ★65%

WI 57.6
2026.3.1
🧠 medium

■42% ▼56% ∅58% ▲58% ★74%

Claude Opus 4.7

WI 71.9k=2
🧠 off

■67% ▼71% ∅72% ▲73% ★76%

WI 72.3k=4
2.1.112
🧠 xhigh

■60% ▼73% ∅73% ▲74% ★82%

WI 74.4k=4
2026.3.11
🧠 off

■56% ▼67% ∅75% ▲81% ★91%

Gemini 3.5 Flash

WI 71.2
🧠 high

■53% ▼69% ∅71% ▲75% ★89%

WI 64.4k=4
🧠 low

■49% ▼61% ∅65% ▲70% ★78%

WI 69.4
v2026.5.16
🧠 high

■49% ▼69% ∅70% ▲72% ★87%

WI 69.2k=3
v2026.5.16
🧠 medium

■54% ▼67% ∅69% ▲71% ★84%

WI 63.7
v2026.5.16
🧠 low

■45% ▼58% ∅64% ▲67% ★82%

WI 72.6k=4
2026.4.23
🧠 high

■56% ▼65% ∅72% ▲81% ★90%

WI 69.1k=4
2026.4.23
🧠 medium

■54% ▼66% ∅69% ▲73% ★83%

WI 67.2k=3
2026.4.23
🧠 low

■55% ▼66% ∅67% ▲69% ★79%

GLM-5.2 [W&B Dynamo]

WI 70.7

■54% ▼70% ∅71% ▲73% ★85%

GLM-5.2 [Z.ai]

WI 71.6
🧠 max

■61% ▼70% ∅71% ▲72% ★83%

WI 67.2
🧠 high

■47% ▼65% ∅69% ▲73% ★83%

WI 64.9
v2026.6.19
🧠 max

■46% ▼62% ∅65% ▲70% ★82%

GPT-5.4

WI 69.9k=3
🧠 xhigh

■55% ▼70% ∅71% ▲73% ★81%

WI 43.3k=2
🧠 off

■33% ▼43% ∅43% ▲44% ★54%

WI 67.0k=3
v2026.3.30
🧠 medium

■54% ▼65% ∅67% ▲71% ★79%

WI 70.0
2026.3.11
🧠 xhigh

■52% ▼70% ∅71% ▲72% ★85%

WI 61.5k=4
2026.3.11
🧠 low

■47% ▼57% ∅61% ▲66% ★76%

WI 30.3
2026.3.1
🧠 off

■9% ▼28% ∅30% ▲33% ★52%

GLM-5.2 [W&B]

WI 69.9k=4

■60% ▼69% ∅70% ▲73% ★80%

WI 69.1k=4
v2026.6.19

■53% ▼67% ∅70% ▲73% ★82%

Claude Fable 5 – 13 refusals

WI N/A
🧠 high

■55% ▼63% ∅68% ▲72% ★78%

Claude Sonnet 5

WI 67.1
🧠 max

■55% ▼65% ∅68% ▲70% ★78%

WI 74.2k=1
2.1.197

74%

WI 61.8k=1
2026.3.1
🧠 high

62%

Claude Sonnet 4.6

WI 61.8

■42% ▼60% ∅62% ▲64% ★81%

WI 57.7k=4
2.1.63

■43% ▼54% ∅58% ▲63% ★72%

WI 53.0k=4
2026.3.1

■40% ▼48% ∅53% ▲56% ★66%

Qwen3.8-27B [W&B]

WI 60.8
🧠 xhigh

■42% ▼58% ∅61% ▲63% ★79%

WI N/A
🧠 medium

■43% ▼55% ∅60% ▲64% ★75%

Kimi K2.6 [Moonshot AI]

WI 59.1k=4

■43% ▼57% ∅60% ▲63% ★73%

WI N/A
v2026.3.30

■28% ▼37% ∅53% ▲64% ★71%

WI 58.4k=2
2026.3.11

■51% ▼56% ∅58% ▲61% ★66%

Kimi K2.7 Code [W&B]

WI 56.9
🧠 off

■38% ▼54% ∅58% ▲61% ★73%

DeepSeek-V4-Pro [W&B]

WI 55.1k=1

55%

WI 38.2k=2
v2026.3.30

■28% ▼35% ∅38% ▲42% ★48%

Kimi K2.6 [W&B]

WI 53.9k=1

54%

WI 54.6
v2026.3.30

■33% ▼48% ∅56% ▲62% ★73%

Gemini 3.1 Pro Preview

WI 53.1k=3

■40% ▼51% ∅53% ▲56% ★65%

WI 57.3k=3
2026.3.11

■45% ▼56% ∅58% ▲60% ★69%

MiniMax M2.7

WI 51.9k=3

■39% ▼49% ∅52% ▲55% ★63%

WI 49.4k=1
2026.3.11

49%

NVIDIA-Nemotron-3-Ultra-550B-A55B [W&B]

WI 49.3

■31% ▼48% ∅49% ▲52% ★66%

WI 41.1
v2026.5.16

■24% ▼40% ∅42% ▲44% ★56%

Kimi K2.5 (int4) [W&B]

WI 46.8k=3

■36% ▼46% ∅47% ▲48% ★57%

WI 39.9k=2
2026.3.1

■30% ▼35% ∅40% ▲45% ★49%

Kimi K2.5 (nvfp4) [W&B]

WI 46.6k=2

■39% ▼46% ∅47% ▲47% ★54%

WI 44.9k=1
v2026.3.30

45%

WI 36.8k=4
2026.3.1

■16% ▼34% ∅37% ▲38% ★58%

GLM-5-Turbo

WI 46.1k=1

46%

WI 46.9k=3
2026.3.11

■33% ▼46% ∅47% ▲47% ★62%

Gemini 3 Flash Preview

WI 42.7k=2

■36% ▼42% ∅43% ▲44% ★49%

GLM-5.1 [W&B]

WI 40.4k=2

■34% ▼39% ∅40% ▲42% ★47%

WI 28.7k=2
2026.3.11

■19% ▼26% ∅29% ▲31% ★38%

GPT-5.3-Codex

WI 38.8k=4

■25% ▼38% ∅39% ▲42% ★52%

WI 54.6k=3
2026.3.1

■40% ▼54% ∅55% ▲56% ★69%

NVIDIA-Nemotron-3-Super-120B-A12B-FP8 [W&B]

WI 35.6

■18% ▼31% ∅36% ▲39% ★53%

WI 20.2k=2
2026.3.1

■12% ▼17% ∅20% ▲24% ★28%

Gemma 4 31B [W&B]

WI 30.3k=1

30%

WI 17.3k=3
2026.3.11

■9% ▼17% ∅18% ▲18% ★25%

GPT‑5.4 mini

WI 26.4k=2

■20% ▼26% ∅26% ▲27% ★33%

WI 15.0
2026.3.11

■4% ▼10% ∅14% ▲18% ★28%

Gemini 3.1 Flash Lite Preview

WI 25.8k=2

■18% ▼26% ∅26% ▲26% ★34%

WI 23.2k=3
2026.3.11

■12% ▼20% ∅23% ▲26% ★35%

GPT‑5.4 nano

WI 22.6k=3

■9% ▼20% ∅22% ▲24% ★37%

WI 14.6k=3
2026.3.11

■7% ▼12% ∅14% ▲17% ★24%

Mistral Small 4 119B A6B

WI 22.5k=2

■18% ▼21% ∅22% ▲24% ★27%

WI 16.3k=2
2026.3.11

■11% ▼16% ∅16% ▲17% ★21%

gpt-6-astra

WI N/A
v2026.8.31
🧠 max

■76% ▼81% ∅83% ▲85% ★89%

WI 85.1
0.153.3
🧠 max

■75% ▼84% ∅86% ▲89% ★92%

DeepSeek-V4-Flash [W&B]

WI 41.6k=1
v2026.3.30

42%

Run Details (327 runs)

Across these 327 runs, 88 (99%) of the 89 tasks were solved at least once, 0 (0%) were solved every time, and 1 (1%) were never solved.

Total runtime: 22d 18h 33m (timing data for 321/327 runs) · Total tokens: 36.23B (35.62B in, 27.26B cached, 618.4M out) (token data for 288/327 runs) · Total cost: ~$33,000.16–$33,026.47 (5 runs use estimated ranges; cost data for 288/327 runs).

Date	Agent	Provider	Vendor	Model	Think	Score	Pass	Fail	Timeout	Timeouts	Duration	In	Out	Total	Cost
2026-09-08 08:13	Hermes Agent (v2026.8.31)	openai	openai	gpt-6-astra	max	85.4%	76	13	3600s	1	1h09m	-	-	-	-
2026-09-08 07:00	Hermes Agent (v2026.8.31)	openai	openai	gpt-6-astra	max	83.1%	74	15	3600s	1	1h12m	-	-	-	-
2026-09-08 05:51	Hermes Agent (v2026.8.31)	openai	openai	gpt-6-astra	max	85.4%	76	13	3600s	1	1h09m	-	-	-	-
2026-09-08 04:42	Hermes Agent (v2026.8.31)	openai	openai	gpt-6-astra	max	82.0%	73	16	3600s	1	1h08m	-	-	-	-
2026-09-07 22:16	Hermes Agent (v2026.8.31)	openai	openai	gpt-6-astra	max	80.9%	72	17	3600s	0	0h59m	-	-	-	-
2026-09-05 17:55	Codex (0.153.3)	openai	openai	gpt-6-astra	max	88.8%	79	10	3600s	0	-	54.1M	1.2M	55.3M	$162.72
2026-09-05 17:22	Codex (0.153.3)	openai	openai	gpt-6-astra	max	86.5%	77	12	3600s	0	-	50.4M	1.1M	51.5M	$155.77
2026-09-05 16:51	Codex (0.153.3)	openai	openai	gpt-6-astra	max	85.4%	76	13	3600s	0	-	54.5M	1.2M	55.7M	$162.94
2026-09-05 16:09	Codex (0.153.3)	openai	openai	gpt-6-astra	max	84.3%	75	14	3600s	0	-	56.9M	1.2M	58.1M	$169.96
2026-09-05 00:42	Codex (0.153.3)	openai	openai	gpt-6-astra	max	84.3%	75	14	3600s	0	-	54.8M	1.2M	56.0M	$165.72
2026-09-02 03:32	Terminus-2 (2.0.0)	wandb	Qwen	Qwen3.8-27B [W&B]	medium	60.7%	54	35	3600s	13	1h11m	93.4M	2.8M	96.2M	$24.12
2026-09-01 07:57	Terminus-2 (2.0.0)	wandb	Qwen	Qwen3.8-27B [W&B]	medium	61.8%	55	34	3600s	9	0h57m	72.3M	2.7M	75.0M	$21.26
2026-09-01 06:19	Terminus-2 (2.0.0)	wandb	Qwen	Qwen3.8-27B [W&B]	medium	64.0%	57	32	3600s	13	1h38m	92.5M	3.0M	95.5M	$25.41
2026-09-01 04:28	Terminus-2 (2.0.0)	wandb	Qwen	Qwen3.8-27B [W&B]	medium	56.2%	50	39	3600s	12	1h50m	103.8M	2.9M	106.7M	$26.63
2026-09-01 02:58	Terminus-2 (2.0.0)	wandb	Qwen	Qwen3.8-27B [W&B]	medium	55.1%	49	40	3600s	15	1h30m	103.1M	3.0M	106.1M	$26.77
2026-08-22 04:46	Terminus-2 (2.0.0)	wandb	Qwen	Qwen3.8-27B [W&B]	xhigh	58.4%	52	37	3600s	27	2h01m	75.2M	2.7M	77.9M	$24.72
2026-08-22 02:35	Terminus-2 (2.0.0)	wandb	Qwen	Qwen3.8-27B [W&B]	xhigh	62.9%	56	33	3600s	23	2h10m	97.2M	3.2M	100.4M	$29.62
2026-08-22 00:29	Terminus-2 (2.0.0)	wandb	Qwen	Qwen3.8-27B [W&B]	xhigh	60.7%	54	35	3600s	20	2h06m	100.6M	3.1M	103.8M	$28.91
2026-08-21 00:21	Terminus-2 (2.0.0)	wandb	Qwen	Qwen3.8-27B [W&B]	xhigh	62.9%	56	33	3600s	14	-	133.0M	4.0M	137.0M	$34.45
2026-08-20 01:19	Terminus-2 (2.0.0)	wandb	Qwen	Qwen3.8-27B [W&B]	xhigh	61.8%	55	34	3600s	16	1h47m	145.0M	4.3M	149.3M	$37.96
2026-07-16 01:22	Codex (0.144.4)	openai	openai	GPT-5.6 Sol	max	83.1%	74	15	3600s	0	15h56m	95.8M	1.1M	96.9M	~$99.38–$104.76
2026-07-15 23:17	Codex (0.144.4)	openai	openai	GPT-5.6 Sol	max	88.8%	79	10	3600s	0	2h00m	96.4M	1.1M	97.5M	~$103.50–$109.69
2026-07-15 21:11	Codex (0.144.4)	openai	openai	GPT-5.6 Sol	max	87.6%	78	11	3600s	0	2h05m	58.4M	806K	59.2M	~$68.98–$73.32
2026-07-15 17:42	Codex (0.144.4)	openai	openai	GPT-5.6 Sol	max	88.8%	79	10	3600s	0	2h07m	79.1M	933K	80.1M	~$86.68–$91.99
2026-07-15 12:12	Codex (0.144.4)	openai	openai	GPT-5.6 Sol	max	85.4%	76	13	3600s	0	1h51m	84.4M	995K	85.4M	~$90.42–$95.52
2026-07-13 11:58	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Luna	max	76.4%	68	21	3600s	2	1h46m	-	-	-	-
2026-07-13 10:17	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Luna	max	80.9%	72	17	3600s	0	1h41m	-	-	-	-
2026-07-13 08:33	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Luna	max	76.4%	68	21	3600s	2	1h43m	-	-	-	-
2026-07-13 06:42	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Luna	max	75.3%	67	22	3600s	3	1h51m	-	-	-	-
2026-07-13 04:58	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Luna	max	84.3%	75	14	3600s	2	1h43m	-	-	-	-
2026-07-13 03:23	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Terra	max	80.9%	72	17	3600s	1	1h34m	-	-	-	-
2026-07-13 01:35	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Terra	max	79.8%	71	18	3600s	3	1h48m	-	-	-	-
2026-07-12 23:48	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Terra	max	79.8%	71	18	3600s	3	1h46m	-	-	-	-
2026-07-12 22:02	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Terra	max	79.8%	71	18	3600s	3	1h46m	-	-	-	-
2026-07-12 20:11	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Terra	max	84.3%	75	14	3600s	2	1h50m	-	-	-	-
2026-07-12 18:21	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Sol	max	85.4%	76	13	3600s	2	1h49m	-	-	-	-
2026-07-12 16:46	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Sol	max	86.5%	77	12	3600s	0	1h34m	-	-	-	-
2026-07-12 15:24	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Sol	max	77.5%	69	20	3600s	0	1h22m	-	-	-	-
2026-07-12 14:02	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Sol	max	88.8%	79	10	3600s	0	1h21m	-	-	-	-
2026-07-12 03:21	Hermes Agent (v2026.7.7.2)	openai	openai	GPT-5.6 Sol	max	84.3%	75	14	3600s	0	1h29m	-	-	-	-
2026-07-11 12:45	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Luna	max	77.5%	69	20	3600s	1	1h11m	74.8M	2.0M	76.8M	$26.44
2026-07-11 11:58	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Luna	max	77.5%	69	20	3600s	0	0h46m	107.3M	2.3M	109.6M	$31.95
2026-07-11 10:42	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Luna	max	78.7%	70	19	3600s	1	1h15m	102.0M	2.2M	104.2M	$30.10
2026-07-11 09:34	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Luna	max	74.2%	66	23	3600s	0	1h07m	112.3M	2.3M	114.6M	$32.08
2026-07-11 08:05	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Luna	max	76.4%	68	21	3600s	2	1h29m	116.0M	2.5M	118.5M	$34.45
2026-07-11 06:42	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Terra	max	79.8%	71	18	3600s	2	1h22m	64.8M	2.5M	67.3M	$66.34
2026-07-11 05:26	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Terra	max	78.7%	70	19	3600s	2	1h15m	59.3M	2.3M	61.6M	$62.28
2026-07-11 04:13	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Terra	max	86.5%	77	12	3600s	2	1h12m	57.2M	2.2M	59.4M	$60.04
2026-07-11 02:59	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Terra	max	79.8%	71	18	3600s	1	1h13m	59.8M	2.3M	62.1M	$61.96
2026-07-11 01:43	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Terra	max	85.4%	76	13	3600s	2	1h15m	52.5M	2.3M	54.8M	$58.69
2026-07-10 07:18	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Sol	max	86.5%	77	12	3600s	4	1h33m	27.4M	1.4M	28.8M	$70.55
2026-07-10 06:01	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Sol	max	84.3%	75	14	3600s	2	1h16m	27.9M	1.4M	29.3M	$72.78
2026-07-10 04:46	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Sol	max	83.1%	74	15	3600s	3	1h14m	32.1M	1.4M	33.5M	$75.08
2026-07-10 03:31	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Sol	max	87.6%	78	11	3600s	2	1h15m	29.2M	1.4M	30.6M	$72.00
2026-07-10 02:16	Terminus-2 (2.0.0)	openai	openai	GPT-5.6 Sol	max	84.3%	75	14	3600s	1	1h14m	32.1M	1.4M	33.5M	$75.34
2026-07-07 08:34	Terminus-2 (2.0.0)	wandb	zai-org	GLM-5.2 [W&B Dynamo]	-	70.8%	63	26	3600s	5	1h56m	80.4M	7.5M	87.9M	$85.89
2026-07-07 06:51	Terminus-2 (2.0.0)	wandb	zai-org	GLM-5.2 [W&B Dynamo]	-	73.0%	65	24	3600s	6	1h43m	112.7M	7.6M	120.3M	$100.25
2026-07-07 05:07	Terminus-2 (2.0.0)	wandb	zai-org	GLM-5.2 [W&B Dynamo]	-	69.7%	62	27	3600s	4	1h43m	98.8M	6.9M	105.7M	$90.34
2026-07-07 03:15	Terminus-2 (2.0.0)	wandb	zai-org	GLM-5.2 [W&B Dynamo]	-	73.0%	65	24	3600s	3	1h51m	74.2M	6.3M	80.5M	$77.83
2026-07-07 01:51	Terminus-2 (2.0.0)	wandb	zai-org	GLM-5.2 [W&B Dynamo]	-	70.8%	63	26	3600s	2	1h23m	69.7M	6.4M	76.1M	$76.72
2026-07-01 18:05	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Sonnet 5	max	67.4%	60	29	3600s	18	2h08m	77.1M	8.9M	85.9M	$114.33
2026-07-01 15:59	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Sonnet 5	max	67.4%	60	29	3600s	17	2h05m	92.7M	8.6M	101.3M	$115.44
2026-07-01 13:47	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Sonnet 5	max	65.2%	58	31	3600s	18	2h11m	94.2M	9.2M	103.4M	$121.20
2026-07-01 11:39	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Sonnet 5	max	69.7%	62	27	3600s	18	2h08m	87.5M	8.9M	96.4M	$117.67
2026-07-01 09:31	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Sonnet 5	max	68.5%	61	28	3600s	18	2h07m	86.6M	8.9M	95.5M	$116.26
2026-06-30 21:46	OpenClaw (2026.3.1)	anthropic	anthropic	Claude Sonnet 5	high	61.8%	55	34	3600s	9	1h28m	-	-	-	-
2026-06-30 21:46	Claude Code (2.1.197)	anthropic	anthropic	Claude Sonnet 5	-	74.2%	66	23	3600s	4	1h04m	234.4M	2.7M	237.1M	$89.32
2026-06-26 20:05	Hermes Agent (v2026.6.19)	openrouter	z-ai	GLM-5.2 [Z.ai]	max	66.3%	59	30	3600s	1	1h38m	-	-	-	-
2026-06-26 18:32	Hermes Agent (v2026.6.19)	openrouter	z-ai	GLM-5.2 [Z.ai]	max	65.2%	58	31	3600s	1	1h32m	-	-	-	-
2026-06-26 17:03	Hermes Agent (v2026.6.19)	openrouter	z-ai	GLM-5.2 [Z.ai]	max	69.7%	62	27	3600s	1	1h28m	-	-	-	-
2026-06-26 15:13	Hermes Agent (v2026.6.19)	openrouter	z-ai	GLM-5.2 [Z.ai]	max	61.8%	55	34	3600s	3	1h49m	-	-	-	-
2026-06-26 13:42	Hermes Agent (v2026.6.19)	openrouter	z-ai	GLM-5.2 [Z.ai]	max	64.0%	57	32	3600s	4	1h30m	-	-	-	-
2026-06-26 09:06	Hermes Agent (v2026.6.19)	wandb	zai-org	GLM-5.2 [W&B]	-	68.5%	61	28	3600s	3	1h39m	-	-	-	-
2026-06-26 05:07	Hermes Agent (v2026.6.19)	wandb	zai-org	GLM-5.2 [W&B]	-	67.4%	60	29	3600s	2	1h32m	-	-	-	-
2026-06-26 03:28	Hermes Agent (v2026.6.19)	wandb	zai-org	GLM-5.2 [W&B]	-	71.9%	64	25	3600s	1	1h39m	-	-	-	-
2026-06-26 02:04	Hermes Agent (v2026.6.19)	wandb	zai-org	GLM-5.2 [W&B]	-	73.0%	65	24	3600s	1	1h22m	-	-	-	-
2026-06-25 01:14	Terminus-2 (2.0.0)	openrouter	z-ai	GLM-5.2 [Z.ai]	max	71.9%	64	25	3600s	14	3h13m	37.4M	4.8M	42.1M	$34.72
2026-06-24 22:19	Terminus-2 (2.0.0)	openrouter	z-ai	GLM-5.2 [Z.ai]	max	71.9%	64	25	3600s	14	2h54m	49.1M	5.6M	54.7M	$40.86
2026-06-24 19:33	Terminus-2 (2.0.0)	openrouter	z-ai	GLM-5.2 [Z.ai]	max	71.9%	64	25	3600s	12	2h45m	70.3M	6.1M	76.3M	$49.35
2026-06-24 16:58	Terminus-2 (2.0.0)	openrouter	z-ai	GLM-5.2 [Z.ai]	max	71.9%	64	25	3600s	10	2h34m	44.1M	5.2M	49.4M	$38.27
2026-06-24 13:48	Terminus-2 (2.0.0)	openrouter	z-ai	GLM-5.2 [Z.ai]	max	69.7%	62	27	3600s	15	3h09m	40.4M	4.9M	45.3M	$36.05
2026-06-24 06:22	Terminus-2 (2.0.0)	wandb	zai-org	GLM-5.2 [W&B]	-	73.0%	65	24	3600s	9	1h43m	70.6M	7.0M	77.6M	$86.17
2026-06-24 04:39	Terminus-2 (2.0.0)	wandb	zai-org	GLM-5.2 [W&B]	-	68.5%	61	28	3600s	8	1h42m	58.2M	7.0M	65.1M	$77.86
2026-06-24 03:02	Terminus-2 (2.0.0)	wandb	zai-org	GLM-5.2 [W&B]	-	69.7%	62	27	3600s	8	1h37m	59.3M	6.0M	65.3M	$77.49
2026-06-23 18:31	Terminus-2 (2.0.0)	wandb	zai-org	GLM-5.2 [W&B]	-	68.5%	61	28	3600s	7	1h42m	64.3M	6.2M	70.6M	$75.13
2026-06-23 00:34	Terminus-2 (2.0.0)	openrouter	z-ai	GLM-5.2 [Z.ai]	high	69.7%	62	27	3600s	6	2h08m	55.3M	3.5M	58.8M	$33.79
2026-06-22 00:32	Terminus-2 (2.0.0)	openrouter	z-ai	GLM-5.2 [Z.ai]	high	65.2%	58	31	3600s	11	1h47m	59.7M	4.0M	63.7M	$37.45
2026-06-21 22:29	Terminus-2 (2.0.0)	openrouter	z-ai	GLM-5.2 [Z.ai]	high	68.5%	61	28	3600s	10	2h02m	61.9M	4.0M	65.9M	$38.23
2026-06-21 20:24	Terminus-2 (2.0.0)	openrouter	z-ai	GLM-5.2 [Z.ai]	high	73.0%	65	24	3600s	4	1h46m	48.8M	3.8M	52.6M	$32.72
2026-06-21 12:13	Terminus-2 (2.0.0)	openrouter	z-ai	GLM-5.2 [Z.ai]	high	66.3%	59	30	3600s	13	2h18m	73.0M	4.1M	77.2M	$42.53
2026-06-17 12:30	Codex (0.125.0)	openai	openai	GPT-5.5	high	78.7%	70	19	3600s	1	3h17m	116.8M	927K	117.7M	$114.02
2026-06-15 14:43	Terminus-2 (2.0.0)	openai	moonshotai	Kimi K2.7 Code [W&B]	off	59.6%	53	36	3600s	21	3h19m	59.5M	2.1M	61.6M	$37.45
2026-06-15 11:46	Terminus-2 (2.0.0)	openai	moonshotai	Kimi K2.7 Code [W&B]	off	57.3%	51	38	3600s	16	2h56m	51.1M	2.0M	53.1M	$32.47
2026-06-15 08:20	Terminus-2 (2.0.0)	openai	moonshotai	Kimi K2.7 Code [W&B]	off	53.9%	48	41	3600s	26	3h26m	47.7M	2.2M	50.0M	$33.61
2026-06-15 05:18	Terminus-2 (2.0.0)	openai	moonshotai	Kimi K2.7 Code [W&B]	off	60.7%	54	35	3600s	19	3h00m	49.8M	2.0M	51.8M	$32.40
2026-06-15 01:59	Terminus-2 (2.0.0)	openai	moonshotai	Kimi K2.7 Code [W&B]	off	57.3%	51	38	3600s	20	3h19m	67.0M	2.2M	69.2M	$37.56
2026-06-10 23:07	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Fable 5 – 13 refusals	high	62.9%	56	33	3600s	17	1h35m	201.1M	1.5M	202.6M	$2,425.18
2026-06-10 20:20	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Fable 5 – 13 refusals	high	66.3%	59	30	3600s	15	1h21m	207.4M	1.4M	208.8M	$2,536.92
2026-06-10 17:21	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Fable 5 – 13 refusals	high	69.7%	62	27	3600s	17	1h24m	210.7M	1.5M	212.2M	$2,573.59
2026-06-10 13:41	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Fable 5 – 13 refusals	high	71.9%	64	25	3600s	16	1h24m	82.8M	1.4M	84.2M	$968.05
2026-06-10 02:21	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Fable 5 – 13 refusals	high	67.4%	60	29	3600s	16	1h40m	212.9M	1.3M	214.3M	$2,577.38
2026-06-05 13:31	Hermes Agent (v2026.5.16)	wandb	nvidia	NVIDIA-Nemotron-3-Ultra-550B-A55B [W&B]	-	41.6%	37	52	3600s	7	1h37m	188.6M	2.1M	190.7M	$147.18
2026-06-05 11:51	Hermes Agent (v2026.5.16)	wandb	nvidia	NVIDIA-Nemotron-3-Ultra-550B-A55B [W&B]	-	40.4%	36	53	3600s	6	1h40m	-	-	-	-
2026-06-05 10:03	Hermes Agent (v2026.5.16)	wandb	nvidia	NVIDIA-Nemotron-3-Ultra-550B-A55B [W&B]	-	43.8%	39	50	3600s	9	1h46m	228.4M	2.3M	230.7M	$177.61
2026-06-05 08:22	Hermes Agent (v2026.5.16)	wandb	nvidia	NVIDIA-Nemotron-3-Ultra-550B-A55B [W&B]	-	40.4%	36	53	3600s	7	1h41m	185.7M	2.1M	187.8M	$145.02
2026-06-05 06:32	Hermes Agent (v2026.5.16)	wandb	nvidia	NVIDIA-Nemotron-3-Ultra-550B-A55B [W&B]	-	42.7%	38	51	3600s	17	1h50m	160.2M	1.8M	162.1M	$125.26
2026-06-05 05:16	Terminus-2 (2.0.0)	wandb	nvidia	NVIDIA-Nemotron-3-Ultra-550B-A55B [W&B]	-	50.6%	45	44	3600s	7	1h15m	140.6M	3.2M	143.8M	$114.27
2026-06-05 03:59	Terminus-2 (2.0.0)	wandb	nvidia	NVIDIA-Nemotron-3-Ultra-550B-A55B [W&B]	-	48.3%	43	46	3600s	8	1h16m	111.7M	3.0M	114.6M	$91.95
2026-06-05 02:46	Terminus-2 (2.0.0)	wandb	nvidia	NVIDIA-Nemotron-3-Ultra-550B-A55B [W&B]	-	48.3%	43	46	3600s	11	1h13m	140.0M	3.5M	143.5M	$114.60
2026-06-05 01:31	Terminus-2 (2.0.0)	wandb	nvidia	NVIDIA-Nemotron-3-Ultra-550B-A55B [W&B]	-	51.7%	46	43	3600s	8	1h14m	103.5M	3.1M	106.7M	$86.31
2026-06-04 23:52	Terminus-2 (2.0.0)	wandb	nvidia	NVIDIA-Nemotron-3-Ultra-550B-A55B [W&B]	-	48.3%	43	46	3600s	10	1h39m	139.0M	3.7M	142.7M	$114.50
2026-05-27 22:50	OpenClaw (2026.4.23)	google	google	Gemini 3.5 Flash	low	66.3%	59	30	3600s	4	2h37m	273.7M	2.4M	276.1M	$97.74
2026-05-27 19:09	OpenClaw (2026.4.23)	google	google	Gemini 3.5 Flash	low	68.5%	61	28	3600s	5	3h40m	350.8M	2.7M	353.5M	$117.91
2026-05-27 13:11	OpenClaw (2026.4.23)	google	google	Gemini 3.5 Flash	low	67.4%	60	29	3600s	7	3h03m	465.1M	2.9M	468.1M	$151.39
2026-05-27 12:05	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3.5 Flash	low	66.3%	59	30	3600s	2	1h06m	339.1M	2.1M	341.2M	$109.98
2026-05-27 10:59	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3.5 Flash	low	62.9%	56	33	3600s	1	1h06m	215.3M	1.8M	217.1M	$84.32
2026-05-27 09:58	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3.5 Flash	low	69.7%	62	27	3600s	2	1h01m	263.6M	2.1M	265.6M	$97.72
2026-05-27 07:50	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3.5 Flash	low	60.7%	54	35	3600s	4	1h01m	378.3M	2.2M	380.5M	$120.31
2026-05-27 06:24	Hermes Agent (v2026.5.16)	google	google	Gemini 3.5 Flash	low	65.2%	58	31	3600s	0	1h25m	132.7M	625K	133.3M	$57.54
2026-05-27 05:16	Hermes Agent (v2026.5.16)	google	google	Gemini 3.5 Flash	low	58.4%	52	37	3600s	0	1h08m	119.6M	555K	120.2M	$53.56
2026-05-27 03:56	Hermes Agent (v2026.5.16)	google	google	Gemini 3.5 Flash	low	67.4%	60	29	3600s	0	1h19m	136.5M	504K	137.0M	$55.11
2026-05-27 02:44	Hermes Agent (v2026.5.16)	google	google	Gemini 3.5 Flash	low	62.9%	56	33	3600s	0	1h12m	148.3M	597K	148.9M	$60.59
2026-05-27 01:35	Hermes Agent (v2026.5.16)	google	google	Gemini 3.5 Flash	low	66.3%	59	30	3600s	0	1h08m	122.1M	553K	122.7M	$56.83
2026-05-26 13:31	OpenClaw (2026.4.23)	google	google	Gemini 3.5 Flash	high	70.8%	63	26	3600s	4	2h28m	328.4M	2.5M	331.0M	$122.12
2026-05-26 06:45	OpenClaw (2026.4.23)	google	google	Gemini 3.5 Flash	high	80.9%	72	17	3600s	8	3h40m	344.2M	2.3M	346.5M	$125.00
2026-05-26 04:22	OpenClaw (2026.4.23)	google	google	Gemini 3.5 Flash	high	71.9%	64	25	3600s	5	2h22m	256.5M	2.2M	258.7M	$91.14
2026-05-26 00:39	OpenClaw (2026.4.23)	google	google	Gemini 3.5 Flash	high	65.2%	58	31	3600s	5	3h42m	416.3M	2.5M	418.8M	$132.44
2026-05-25 09:27	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3.5 Flash	high	68.5%	61	28	3600s	1	1h06m	287.3M	2.9M	290.2M	$114.19
2026-05-25 08:22	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3.5 Flash	high	69.7%	62	27	3600s	4	1h05m	421.7M	3.9M	425.5M	$154.30
2026-05-25 07:15	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3.5 Flash	high	69.7%	62	27	3600s	2	1h06m	259.6M	3.1M	262.7M	$112.24
2026-05-25 05:40	Hermes Agent (v2026.5.16)	google	google	Gemini 3.5 Flash	high	69.7%	62	27	3600s	0	1h35m	176.5M	751K	177.3M	$69.57
2026-05-25 04:13	Hermes Agent (v2026.5.16)	google	google	Gemini 3.5 Flash	high	70.8%	63	26	3600s	0	1h27m	190.3M	728K	191.0M	$75.54
2026-05-25 02:55	Hermes Agent (v2026.5.16)	google	google	Gemini 3.5 Flash	high	70.8%	63	26	3600s	0	1h17m	178.1M	774K	178.8M	$75.25
2026-05-25 01:28	Hermes Agent (v2026.5.16)	google	google	Gemini 3.5 Flash	high	68.5%	61	28	3600s	0	1h27m	-	-	-	-
2026-05-25 00:56	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3.5 Flash	high	74.2%	66	23	3600s	0	0h31m	214.5M	2.8M	217.3M	$95.44
2026-05-24 16:15	Hermes Agent (v2026.5.16)	google	google	Gemini 3.5 Flash	high	71.9%	64	25	3600s	0	1h33m	186.1M	724K	186.8M	$72.26
2026-05-24 15:09	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3.5 Flash	high	75.3%	67	22	3600s	3	1h06m	449.5M	3.0M	452.5M	$144.63
2026-05-24 01:13	OpenClaw (2026.4.23)	google	google	Gemini 3.5 Flash	medium	66.3%	59	30	3600s	5	2h13m	293.1M	2.4M	295.5M	$103.64
2026-05-23 19:46	OpenClaw (2026.4.23)	google	google	Gemini 3.5 Flash	medium	73.0%	65	24	3600s	7	2h44m	397.3M	2.9M	400.2M	$140.11
2026-05-23 17:25	OpenClaw (2026.4.23)	google	google	Gemini 3.5 Flash	medium	71.9%	64	25	3600s	6	2h21m	450.2M	2.7M	452.9M	$143.16
2026-05-23 15:04	OpenClaw (2026.4.23)	google	google	Gemini 3.5 Flash	medium	66.3%	59	30	3600s	5	2h20m	340.9M	2.5M	343.3M	$114.63
2026-05-21 07:56	Hermes Agent (v2026.5.16)	google	google	Gemini 3.5 Flash	medium	67.4%	60	29	3600s	0	2h04m	-	-	-	-
2026-05-21 05:55	Hermes Agent (v2026.5.16)	google	google	Gemini 3.5 Flash	medium	70.8%	63	26	3600s	0	2h00m	198.7M	716K	199.5M	$75.99
2026-05-21 01:55	Hermes Agent (v2026.5.16)	google	google	Gemini 3.5 Flash	medium	69.7%	62	27	3600s	0	1h59m	168.0M	715K	168.7M	$69.51
2026-05-16 09:02	Hermes Agent (v2026.3.30)	wandb	moonshotai	Kimi K2.6 [W&B]	-	59.6%	53	36	3600s	5	3h00m	77.8M	5.1M	82.9M	$55.68
2026-05-16 04:31	Hermes Agent (v2026.3.30)	wandb	moonshotai	Kimi K2.6 [W&B]	-	48.3%	43	46	3600s	5	3h29m	77.7M	5.7M	83.4M	$59.54
2026-05-16 03:29	Terminus-2 (2.0.0)	wandb	moonshotai	Kimi K2.6 [W&B]	-	53.9%	48	41	3600s	10	1h01m	135.9M	5.1M	141.0M	$76.43
2026-05-16 00:40	Hermes Agent (v2026.3.30)	wandb	moonshotai	Kimi K2.6 [W&B]	-	60.7%	54	35	3600s	6	2h49m	76.6M	4.9M	81.5M	$54.36
2026-05-15 20:38	Hermes Agent (v2026.3.30)	wandb	moonshotai	Kimi K2.6 [W&B]	-	49.4%	44	45	3600s	4	3h00m	78.8M	4.8M	83.6M	$56.17
2026-05-15 13:10	Hermes Agent (v2026.3.30)	wandb	moonshotai	Kimi K2.6 [W&B]	-	61.8%	55	34	3600s	5	3h08m	84.0M	5.7M	89.8M	$61.07
2026-05-14 03:32	Hermes Agent (v2026.3.30)	wandb	deepseek-ai	DeepSeek-V4-Pro [W&B]	-	41.6%	37	52	3600s	15	3h29m	-	-	-	-
2026-05-14 01:42	Terminus-2 (2.0.0)	wandb	deepseek-ai	DeepSeek-V4-Pro [W&B]	-	55.1%	49	40	3600s	21	1h50m	91.1M	1.9M	93.0M	$94.09
2026-05-13 22:37	Hermes Agent (v2026.3.30)	wandb	deepseek-ai	DeepSeek-V4-Pro [W&B]	-	34.8%	31	58	3600s	6	3h04m	-	-	-	-
2026-05-12 16:25	Hermes Agent (v2026.3.30)	wandb	deepseek-ai	DeepSeek-V4-Flash [W&B]	-	41.6%	37	52	3600s	3	4h15m	70.6M	973K	71.6M	$6.67
2026-04-30 07:51	Codex (0.125.0)	openai	openai	GPT-5.5	high	79.8%	71	18	3600s	1	1h53m	84.7M	833K	85.6M	$214.93
2026-04-30 05:39	Codex (0.125.0)	openai	openai	GPT-5.5	high	77.5%	69	20	3600s	1	2h11m	106.9M	919K	107.8M	$233.29
2026-04-30 04:17	Codex (0.125.0)	openai	openai	GPT-5.5	high	79.8%	71	18	3600s	1	1h21m	115.6M	873K	116.5M	$207.62
2026-04-30 02:51	Codex (0.125.0)	openai	openai	GPT-5.5	high	77.5%	69	20	3600s	1	1h25m	96.3M	850K	97.2M	$198.82
2026-04-27 06:24	Hermes Agent (v2026.4.23)	openai	openai	GPT-5.5	xhigh	73.0%	65	24	3600s	1	1h07m	102.3M	1.2M	103.5M	$116.28
2026-04-27 05:16	Hermes Agent (v2026.4.23)	openai	openai	GPT-5.5	xhigh	77.5%	69	20	3600s	1	1h08m	95.2M	1.2M	96.4M	$110.16
2026-04-27 04:09	Hermes Agent (v2026.4.23)	openai	openai	GPT-5.5	xhigh	71.9%	64	25	3600s	1	1h06m	104.0M	1.2M	105.3M	$120.44
2026-04-27 03:00	Hermes Agent (v2026.4.23)	openai	openai	GPT-5.5	xhigh	74.2%	66	23	3600s	1	1h08m	100.4M	1.3M	101.6M	$115.23
2026-04-27 01:53	Hermes Agent (v2026.4.23)	openai	openai	GPT-5.5	xhigh	74.2%	66	23	3600s	1	1h07m	102.0M	1.2M	103.2M	$126.55
2026-04-26 11:02	Cursor (2026.04.17)	cursor	cursor	GPT-5.5	high	79.8%	71	18	3600s	1	1h06m	57.3M	783K	58.1M	$65.86
2026-04-26 10:17	Cursor (2026.04.17)	cursor	cursor	GPT-5.5	high	74.2%	66	23	3600s	0	0h44m	57.5M	830K	58.3M	$70.17
2026-04-26 07:00	Hermes Agent (v2026.4.23)	openai	openai	GPT-5.5	medium	70.8%	63	26	3600s	1	1h05m	69.5M	626K	70.2M	$89.35
2026-04-26 06:21	Cursor (2026.04.17)	cursor	cursor	GPT-5.5	high	79.8%	71	18	3600s	1	1h56m	51.6M	740K	52.4M	$62.91
2026-04-26 05:55	Hermes Agent (v2026.4.23)	openai	openai	GPT-5.5	medium	70.8%	63	26	3600s	1	1h04m	69.7M	635K	70.4M	$78.26
2026-04-26 04:59	Hermes Agent (v2026.4.23)	openai	openai	GPT-5.5	medium	75.3%	67	22	3600s	0	0h55m	70.4M	617K	71.0M	$91.84
2026-04-26 04:05	Hermes Agent (v2026.4.23)	openai	openai	GPT-5.5	medium	71.9%	64	25	3600s	0	0h53m	63.2M	613K	63.8M	$83.20
2026-04-26 03:09	Hermes Agent (v2026.4.23)	openai	openai	GPT-5.5	medium	75.3%	67	22	3600s	0	0h56m	60.6M	616K	61.3M	$93.92
2026-04-25 21:26	Terminus-2 (2.0.0)	openai	openai	GPT-5.5	xhigh	75.3%	67	22	3600s	5	1h15m	21.9M	2.3M	24.2M	$90.29
2026-04-25 17:15	Terminus-2 (2.0.0)	openai	openai	GPT-5.5	xhigh	78.7%	70	19	3600s	5	1h12m	27.8M	2.4M	30.2M	$106.37
2026-04-25 15:58	Terminus-2 (2.0.0)	openai	openai	GPT-5.5	xhigh	76.4%	68	21	3600s	4	1h16m	20.3M	2.4M	22.7M	$98.74
2026-04-25 12:44	Terminus-2 (2.0.0)	openai	openai	GPT-5.5	medium	68.5%	61	28	3600s	2	1h02m	19.0M	888K	19.9M	$55.56
2026-04-25 11:43	Terminus-2 (2.0.0)	openai	openai	GPT-5.5	medium	71.9%	64	25	3600s	1	1h00m	17.4M	834K	18.3M	$52.43
2026-04-25 08:56	Terminus-2 (2.0.0)	openai	openai	GPT-5.5	medium	67.4%	60	29	3600s	2	1h06m	14.9M	818K	15.8M	$51.27
2026-04-25 07:46	Terminus-2 (2.0.0)	openai	openai	GPT-5.5	medium	69.7%	62	27	3600s	4	1h09m	31.5M	1.0M	32.5M	$67.03
2026-04-25 06:39	OpenClaw (2026.4.23)	openai	openai	GPT-5.5	off	65.2%	58	31	3600s	6	1h06m	65.0M	540K	65.6M	$97.35
2026-04-25 05:32	OpenClaw (2026.4.23)	openai	openai	GPT-5.5	off	70.8%	63	26	3600s	5	1h06m	76.1M	583K	76.6M	$113.39
2026-04-25 04:25	OpenClaw (2026.4.23)	openai	openai	GPT-5.5	off	69.7%	62	27	3600s	6	1h07m	65.6M	556K	66.1M	$118.24
2026-04-25 01:38	OpenClaw (2026.4.23)	openai	openai	GPT-5.5	off	71.9%	64	25	3600s	6	1h06m	78.0M	593K	78.6M	$143.67
2026-04-24 21:18	Hermes Agent (v2026.3.30)	wandbqa	moonshotai	Kimi K2.6 [Moonshot AI]	-	37.1%	33	56	3600s	49	4h32m	-	-	-	-
2026-04-23 10:40	Hermes Agent (v2026.3.30)	moonshotai	moonshotai	Kimi K2.6 [Moonshot AI]	-	57.3%	51	38	3600s	13	3h50m	-	-	-	-
2026-04-23 06:22	Hermes Agent (v2026.3.30)	moonshotai	moonshotai	Kimi K2.6 [Moonshot AI]	-	64.0%	57	32	3600s	14	4h17m	-	-	-	-
2026-04-21 18:27	Terminus-2 (2.0.0)	openrouter	moonshotai	Kimi K2.6 [Moonshot AI]	-	60.7%	54	35	3600s	15	1h55m	60.0M	3.2M	63.1M	$23.91
2026-04-21 14:07	Terminus-2 (2.0.0)	openrouter	moonshotai	Kimi K2.6 [Moonshot AI]	-	62.9%	56	33	3600s	14	1h47m	57.4M	2.8M	60.3M	$22.44
2026-04-21 11:47	Terminus-2 (2.0.0)	openrouter	moonshotai	Kimi K2.6 [Moonshot AI]	-	58.4%	52	37	3600s	13	2h20m	51.2M	2.8M	54.0M	$21.22
2026-04-21 09:59	Terminus-2 (2.0.0)	openrouter	moonshotai	Kimi K2.6 [Moonshot AI]	-	57.3%	51	38	3600s	14	1h47m	55.4M	3.0M	58.4M	$22.61
2026-04-21 05:29	OpenClaw (2026.3.11)	openrouter	moonshotai	Kimi K2.6 [Moonshot AI]	-	56.2%	50	39	3600s	11	1h15m	171.8M	3.4M	175.2M	$55.80
2026-04-21 02:32	OpenClaw (2026.3.11)	openrouter	moonshotai	Kimi K2.6 [Moonshot AI]	-	60.7%	54	35	3600s	11	1h16m	161.6M	3.5M	165.1M	$51.09
2026-04-17 23:30	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Opus 4.7	off	70.8%	63	26	3600s	2	1h03m	57.3M	998K	58.3M	$68.67
2026-04-17 21:52	OpenClaw (2026.3.11)	anthropic	anthropic	Claude Opus 4.7	off	77.5%	69	20	3600s	7	1h38m	243.7M	1.7M	245.4M	$325.96
2026-04-17 20:47	Claude Code (2.1.112)	anthropic	anthropic	Claude Opus 4.7	xhigh	73.0%	65	24	3600s	3	1h05m	114.5M	2.5M	117.1M	$149.08
2026-04-17 19:42	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Opus 4.7	off	73.0%	65	24	3600s	1	1h03m	73.5M	1.0M	74.6M	$77.47
2026-04-17 18:31	OpenClaw (2026.3.11)	anthropic	anthropic	Claude Opus 4.7	off	67.4%	60	29	3600s	4	1h10m	173.9M	1.4M	175.3M	$235.85
2026-04-17 16:52	Claude Code (2.1.112)	anthropic	anthropic	Claude Opus 4.7	xhigh	73.0%	65	24	3600s	6	1h39m	112.3M	2.3M	114.5M	$142.80
2026-04-17 11:48	OpenClaw (2026.3.11)	anthropic	anthropic	Claude Opus 4.7	off	74.2%	66	23	3600s	3	1h04m	154.4M	1.3M	155.7M	$169.95
2026-04-17 10:09	Claude Code (2.1.112)	anthropic	anthropic	Claude Opus 4.7	xhigh	73.0%	65	24	3600s	7	1h39m	114.9M	2.3M	117.2M	$144.82
2026-04-17 02:46	OpenClaw (2026.3.11)	anthropic	anthropic	Claude Opus 4.7	off	80.9%	72	17	3600s	5	1h17m	137.3M	1.2M	138.5M	$142.17
2026-04-17 01:18	Claude Code (2.1.112)	anthropic	anthropic	Claude Opus 4.7	xhigh	74.2%	66	23	3600s	5	1h28m	125.1M	2.4M	127.5M	$157.08
2026-04-15 00:23	OpenClaw (2026.3.11)	wandb	zai-org	GLM-5.1 [W&B]	-	31.5%	28	61	3600s	2	1h04m	24.0M	461K	24.5M	$15.20
2026-04-14 18:51	Terminus-2 (2.0.0)	wandb	zai-org	GLM-5.1 [W&B]	-	41.6%	37	52	3600s	38	2h25m	16.3M	925K	17.3M	$18.26
2026-04-14 07:53	Terminus-2 (2.0.0)	wandb	zai-org	GLM-5.1 [W&B]	-	39.3%	35	54	3600s	44	3h21m	9.6M	642K	10.2M	$12.28
2026-04-13 23:48	OpenClaw (2026.3.11)	wandb	zai-org	GLM-5.1 [W&B]	-	25.8%	23	66	3600s	4	1h31m	22.3M	395K	22.7M	$13.78
2026-04-09 08:03	Terminus-2 (2.0.0)	wandb	google	Gemma 4 31B [W&B]	-	30.3%	27	62	3600s	11	1h27m	127.1M	1.2M	128.3M	$39.63
2026-04-09 03:57	OpenClaw (2026.3.11)	wandb	google	Gemma 4 31B [W&B]	-	18.0%	16	73	3600s	2	1h11m	203.5M	1.5M	205.0M	$62.97
2026-04-09 00:25	OpenClaw (2026.3.11)	wandb	google	Gemma 4 31B [W&B]	-	16.9%	15	74	3600s	3	1h26m	179.6M	1.6M	181.2M	$55.84
2026-04-08 23:18	OpenClaw (2026.3.11)	wandb	google	Gemma 4 31B [W&B]	-	18.0%	16	73	3600s	7	1h07m	124.2M	1.5M	125.7M	$39.12
2026-04-06 05:06	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3.1 Flash Lite Preview	-	25.8%	23	66	3600s	2	1h05m	52.6M	2.0M	54.6M	$6.18
2026-04-06 02:22	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3.1 Flash Lite Preview	-	25.8%	23	66	3600s	2	1h02m	117.4M	2.4M	119.8M	$9.14
2026-04-06 01:18	OpenClaw (2026.3.11)	google	google	Gemini 3.1 Flash Lite Preview	-	20.2%	18	71	3600s	5	1h04m	239.4M	837K	240.3M	$72.40
2026-04-05 21:45	OpenClaw (2026.3.11)	google	google	Gemini 3.1 Flash Lite Preview	-	22.5%	20	69	3600s	4	1h51m	162.3M	741K	163.0M	$51.04
2026-04-05 17:06	OpenClaw (2026.3.11)	google	google	Gemini 3.1 Flash Lite Preview	-	25.8%	23	66	3600s	6	2h57m	123.1M	697K	123.8M	$38.97
2026-04-05 09:59	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3 Flash Preview	-	41.6%	37	52	3600s	5	1h05m	141.3M	1.2M	142.5M	$16.05
2026-04-05 06:09	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3 Flash Preview	-	43.8%	39	50	3600s	5	1h05m	131.9M	1.2M	133.1M	$15.29
2026-04-03 11:44	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3.1 Pro Preview	-	50.6%	45	44	3600s	0	0h31m	9.4M	653K	10.1M	$15.80
2026-04-03 11:08	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3.1 Pro Preview	-	56.2%	50	39	3600s	0	0h35m	9.5M	568K	10.0M	$14.67
2026-04-03 10:41	Terminus-2 (2.0.0)	gemini	gemini	Gemini 3.1 Pro Preview	-	52.8%	47	42	3600s	0	0h26m	13.8M	700K	14.5M	$18.04
2026-04-03 06:53	OpenClaw (2026.3.11)	google	google	Gemini 3.1 Pro Preview	-	59.6%	53	36	3600s	6	1h05m	228.7M	638K	229.3M	$269.13
2026-04-03 05:46	OpenClaw (2026.3.11)	google	google	Gemini 3.1 Pro Preview	-	57.3%	51	38	3600s	5	1h06m	226.5M	748K	227.2M	$266.99
2026-04-03 01:18	OpenClaw (2026.3.11)	google	google	Gemini 3.1 Pro Preview	-	56.2%	50	39	3600s	6	1h06m	102.8M	485K	103.3M	$126.81
2026-04-02 07:08	Hermes Agent (v2026.3.30)	anthropic	anthropic	Claude Opus 4.6	off	64.0%	57	32	3600s	5	4h51m	75.6M	1.0M	76.7M	$86.08
2026-04-02 05:25	Hermes Agent (v2026.3.30)	openai	openai	GPT-5.4	medium	70.8%	63	26	3600s	2	2h38m	80.2M	996K	81.2M	$41.56
2026-04-02 02:55	Hermes Agent (v2026.3.30)	openai	openai	GPT-5.4	medium	65.2%	58	31	3600s	2	2h28m	70.8M	960K	71.8M	$39.07
2026-04-02 00:41	Hermes Agent (v2026.3.30)	anthropic	anthropic	Claude Opus 4.6	off	67.4%	60	29	3600s	5	6h25m	76.1M	1.2M	77.3M	$92.25
2026-04-01 20:09	Hermes Agent (v2026.3.30)	openai	openai	GPT-5.4	medium	66.3%	59	30	3600s	3	2h22m	-	-	-	-
2026-04-01 19:52	Hermes Agent (v2026.3.30)	wandb	moonshotai	Kimi K2.5 (nvfp4) [W&B]	-	44.9%	40	49	3600s	4	3h38m	77.7M	1.6M	79.3M	$19.64
2026-04-01 19:49	Hermes Agent (v2026.3.30)	anthropic	anthropic	Claude Opus 4.6	off	61.8%	55	34	3600s	4	4h51m	69.4M	1.1M	70.5M	$83.38
2026-03-20 06:43	Terminus-2 (2.0.0)	openrouter	minimax	MiniMax M2.7	-	49.4%	44	45	3600s	18	1h33m	126.7M	2.4M	129.1M	$12.49
2026-03-20 02:30	OpenClaw (2026.3.11)	openrouter	minimax	MiniMax M2.7	-	49.4%	44	45	3600s	7	1h14m	135.9M	2.4M	138.3M	$12.69
2026-03-19 13:39	Terminus-2 (2.0.0)	openai	openai	GPT‑5.4 nano	-	20.2%	18	71	3600s	32	2h01m	674.5M	2.6M	677.1M	$21.67
2026-03-19 12:31	Terminus-2 (2.0.0)	openrouter	minimax	MiniMax M2.7	-	52.8%	47	42	3600s	15	1h31m	162.6M	2.7M	165.3M	$19.86
2026-03-19 12:08	Terminus-2 (2.0.0)	openai	openai	GPT‑5.4 nano	-	23.6%	21	68	3600s	28	1h31m	549.2M	2.2M	551.4M	$17.69
2026-03-19 10:43	Terminus-2 (2.0.0)	openai	openai	GPT‑5.4 nano	-	23.6%	21	68	3600s	21	1h24m	550.9M	2.0M	552.9M	$16.52
2026-03-19 09:19	Terminus-2 (2.0.0)	openrouter	minimax	MiniMax M2.7	-	55.1%	49	40	3600s	19	1h30m	99.7M	2.4M	102.2M	$10.72
2026-03-19 09:18	Terminus-2 (2.0.0)	openai	openai	GPT‑5.4 mini	-	27.0%	24	65	3600s	24	1h24m	448.2M	1.3M	449.5M	$73.71
2026-03-19 08:13	Terminus-2 (2.0.0)	mistral	mistral	Mistral Small 4 119B A6B	-	21.3%	19	70	3600s	4	1h15m	242.3M	1.6M	243.9M	$8.54
2026-03-19 06:59	Terminus-2 (2.0.0)	mistral	mistral	Mistral Small 4 119B A6B	-	23.6%	21	68	3600s	4	1h13m	128.5M	1.5M	130.0M	$6.19
2026-03-19 05:57	Terminus-2 (2.0.0)	openai	openai	GPT‑5.4 mini	-	25.8%	23	66	3600s	17	1h39m	450.6M	1.3M	451.9M	$75.65
2026-03-19 04:56	OpenClaw (2026.3.11)	openai	openai	GPT‑5.4 nano	-	12.4%	11	78	3600s	1	1h00m	25.2M	156K	25.4M	$1.00
2026-03-19 04:48	OpenClaw (2026.3.11)	mistral	mistral	Mistral Small 4 119B A6B	-	16.9%	15	74	3600s	6	1h05m	120.9M	842K	121.8M	$4.94
2026-03-19 03:54	OpenClaw (2026.3.11)	openai	openai	GPT‑5.4 nano	-	13.5%	12	77	3600s	1	1h02m	19.4M	143K	19.6M	$0.83
2026-03-19 03:24	OpenClaw (2026.3.11)	mistral	mistral	Mistral Small 4 119B A6B	-	15.7%	14	75	3600s	7	1h23m	115.9M	758K	116.7M	$5.35
2026-03-19 02:53	OpenClaw (2026.3.11)	openai	openai	GPT‑5.4 nano	-	16.9%	15	74	3600s	1	1h00m	13.4M	123K	13.5M	$0.74
2026-03-18 08:05	OpenClaw (2026.3.11)	openai	openai	GPT‑5.4 mini	-	10.1%	9	80	3600s	2	1h02m	20.3M	170K	20.5M	$3.46
2026-03-18 07:03	OpenClaw (2026.3.11)	openai	openai	GPT‑5.4 mini	-	14.6%	13	76	3600s	3	1h02m	16.7M	159K	16.8M	$3.03
2026-03-18 06:01	OpenClaw (2026.3.11)	openai	openai	GPT‑5.4 mini	-	14.6%	13	76	3600s	2	1h02m	21.0M	164K	21.2M	$3.54
2026-03-18 04:58	OpenClaw (2026.3.11)	openai	openai	GPT‑5.4 mini	-	18.0%	16	73	3600s	1	1h02m	19.4M	162K	19.6M	$3.36
2026-03-18 03:56	OpenClaw (2026.3.11)	openai	openai	GPT‑5.4 mini	-	13.5%	12	77	3600s	1	1h02m	16.9M	155K	17.1M	$3.09
2026-03-16 11:53	Terminus-2 (2.0.0)	openrouter	z-ai	GLM-5-Turbo	-	46.1%	41	48	3600s	14	2h03m	198.6M	2.5M	201.2M	$131.84
2026-03-16 10:27	OpenClaw (2026.3.11)	openrouter	z-ai	GLM-5-Turbo	-	47.2%	42	47	3600s	7	1h25m	117.6M	3.4M	121.0M	$61.88
2026-03-16 09:17	OpenClaw (2026.3.11)	openrouter	z-ai	GLM-5-Turbo	-	46.1%	41	48	3600s	6	1h10m	65.6M	2.5M	68.1M	$38.55
2026-03-16 07:45	OpenClaw (2026.3.11)	openrouter	z-ai	GLM-5-Turbo	-	47.2%	42	47	3600s	10	1h31m	72.0M	2.8M	74.8M	$44.98
2026-03-16 01:49	Terminus-2 (2.0.0)	wandb	nvidia	NVIDIA-Nemotron-3-Super-120B-A12B-FP8 [W&B]	-	31.5%	28	61	3600s	21	2h01m	153.9M	4.0M	157.9M	$34.00
2026-03-15 23:55	Terminus-2 (2.0.0)	wandb	nvidia	NVIDIA-Nemotron-3-Super-120B-A12B-FP8 [W&B]	-	38.2%	34	55	3600s	19	1h54m	150.4M	3.9M	154.3M	$33.17
2026-03-15 22:13	Terminus-2 (2.0.0)	wandb	nvidia	NVIDIA-Nemotron-3-Super-120B-A12B-FP8 [W&B]	-	31.5%	28	61	3600s	16	1h41m	132.0M	3.9M	135.9M	$29.52
2026-03-15 20:17	Terminus-2 (2.0.0)	wandb	nvidia	NVIDIA-Nemotron-3-Super-120B-A12B-FP8 [W&B]	-	38.2%	34	55	3600s	19	1h55m	151.3M	4.0M	155.3M	$33.49
2026-03-15 18:06	Terminus-2 (2.0.0)	wandb	nvidia	NVIDIA-Nemotron-3-Super-120B-A12B-FP8 [W&B]	-	39.3%	35	54	3600s	22	2h10m	177.4M	4.3M	181.7M	$38.92
2026-03-14 22:09	OpenClaw (2026.3.11)	anthropic	anthropic	Claude Opus 4.6	max	59.6%	53	36	3600s	8	1h39m	88.2M	1.7M	89.9M	$153.44
2026-03-14 20:31	OpenClaw (2026.3.11)	anthropic	anthropic	Claude Opus 4.6	max	58.4%	52	37	3600s	7	1h37m	104.2M	1.6M	105.8M	$173.63
2026-03-14 18:18	OpenClaw (2026.3.1)	wandb	nvidia	NVIDIA-Nemotron-3-Super-120B-A12B-FP8 [W&B]	-	23.6%	21	68	3600s	7	1h22m	83.9M	1.0M	84.9M	$17.59
2026-03-14 17:09	OpenClaw (2026.3.1)	wandb	nvidia	NVIDIA-Nemotron-3-Super-120B-A12B-FP8 [W&B]	-	16.9%	15	74	3600s	4	1h08m	54.7M	773K	55.4M	$11.55
2026-03-14 07:02	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Opus 4.6	max	60.7%	54	35	3600s	16	1h45m	42.4M	2.3M	44.7M	$93.61
2026-03-14 04:57	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Opus 4.6	max	61.8%	55	34	3600s	15	2h04m	38.8M	2.3M	41.2M	$92.72
2026-03-14 02:54	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Opus 4.6	max	60.7%	54	35	3600s	19	2h02m	38.1M	2.3M	40.4M	$91.68
2026-03-12 20:33	OpenClaw (2026.3.11)	openai	openai	GPT-5.4	low	61.8%	55	34	3600s	11	1h10m	81.0M	613K	81.6M	$49.96
2026-03-12 19:23	OpenClaw (2026.3.11)	openai	openai	GPT-5.4	low	57.3%	51	38	3600s	10	1h10m	70.8M	602K	71.4M	$44.48
2026-03-12 18:18	OpenClaw (2026.3.11)	openai	openai	GPT-5.4	low	59.6%	53	36	3600s	10	1h05m	67.8M	618K	68.4M	$43.76
2026-03-12 17:08	OpenClaw (2026.3.11)	openai	openai	GPT-5.4	low	66.3%	59	30	3600s	10	1h09m	79.2M	603K	79.8M	$48.90
2026-03-12 12:49	OpenClaw (2026.3.11)	openai	openai	GPT-5.4	xhigh	70.8%	63	26	3600s	13	1h12m	141.0M	1.7M	142.7M	$110.57
2026-03-12 11:27	OpenClaw (2026.3.11)	openai	openai	GPT-5.4	xhigh	71.9%	64	25	3600s	11	1h21m	135.4M	1.7M	137.1M	$117.72
2026-03-12 10:16	OpenClaw (2026.3.11)	openai	openai	GPT-5.4	xhigh	70.8%	63	26	3600s	13	1h10m	147.5M	1.8M	149.3M	$108.25
2026-03-12 09:03	OpenClaw (2026.3.11)	openai	openai	GPT-5.4	xhigh	69.7%	62	27	3600s	12	1h12m	156.4M	1.9M	158.3M	$114.24
2026-03-12 08:39	Terminus-2 (2.0.0)	openai	openai	GPT-5.4	xhigh	73.0%	65	24	3600s	11	1h47m	13.5M	6.3M	19.9M	$105.05
2026-03-12 07:38	OpenClaw (2026.3.11)	openai	openai	GPT-5.4	xhigh	71.9%	64	25	3600s	10	1h25m	145.6M	1.6M	147.2M	$104.94
2026-03-12 04:57	Terminus-2 (2.0.0)	openai	openai	GPT-5.4	xhigh	70.8%	63	26	3600s	8	1h19m	10.3M	5.5M	15.8M	$91.64
2026-03-12 03:25	Terminus-2 (2.0.0)	openai	openai	GPT-5.4	xhigh	69.7%	62	27	3600s	11	1h31m	13.3M	5.7M	19.0M	$98.24
2026-03-10 12:31	Terminus-2 (2.0.0)	openai	openai	Kimi K2.5 (nvfp4) [W&B]	-	47.2%	42	47	3600s	12	1h27m	114.4M	1.8M	116.2M	$73.93
2026-03-10 11:30	OpenClaw (2026.3.1)	openai	openai	GPT-5.3-Codex	-	53.9%	48	41	3600s	5	1h04m	33.1M	360K	33.5M	$15.86
2026-03-10 08:42	Terminus-2 (2.0.0)	openai	openai	Kimi K2.5 (nvfp4) [W&B]	-	46.1%	41	48	3600s	13	1h32m	138.5M	2.0M	140.5M	$89.11
2026-03-10 07:38	OpenClaw (2026.3.1)	openai	openai	GPT-5.3-Codex	-	56.2%	50	39	3600s	5	1h04m	31.6M	371K	32.0M	$14.97
2026-03-10 06:29	OpenClaw (2026.3.1)	openai	openai	GPT-5.3-Codex	-	53.9%	48	41	3600s	8	1h09m	30.3M	351K	30.7M	$14.34
2026-03-10 04:58	OpenClaw (2026.3.1)	custom	custom	Kimi K2.5 (nvfp4) [W&B]	-	37.1%	33	56	3600s	14	1h26m	181.2M	1.5M	182.7M	$113.17
2026-03-10 03:33	OpenClaw (2026.3.1)	custom	custom	Kimi K2.5 (nvfp4) [W&B]	-	38.2%	34	55	3600s	10	1h25m	144.8M	1.3M	146.1M	$90.83
2026-03-10 01:37	OpenClaw (2026.3.1)	custom	custom	Kimi K2.5 (nvfp4) [W&B]	-	33.7%	30	59	3600s	9	1h55m	167.6M	1.4M	169.0M	$104.75
2026-03-10 00:20	OpenClaw (2026.3.1)	custom	custom	Kimi K2.5 (nvfp4) [W&B]	-	38.2%	34	55	3600s	10	1h16m	237.4M	1.6M	239.1M	$147.37
2026-03-09 14:05	Terminus-2 (2.0.0)	openai	openai	GPT-5.3-Codex	-	38.2%	34	55	3600s	13	1h14m	233.9M	622K	234.5M	$65.24
2026-03-09 12:59	Terminus-2 (2.0.0)	openai	openai	GPT-5.3-Codex	-	41.6%	37	52	3600s	13	1h05m	243.8M	570K	244.4M	$65.75
2026-03-09 11:33	Terminus-2 (2.0.0)	openai	openai	GPT-5.3-Codex	-	38.2%	34	55	3600s	17	1h25m	320.8M	671K	321.4M	$85.47
2026-03-09 08:09	Terminus-2 (2.0.0)	openai	openai	GPT-5.3-Codex	-	38.2%	34	55	3600s	13	1h32m	245.1M	687K	245.8M	$68.48
2026-03-09 07:46	OpenClaw (2026.3.1)	openai	openai	GPT-5.4	off	28.1%	25	64	3600s	7	1h04m	29.5M	288K	29.8M	$30.21
2026-03-09 06:38	OpenClaw (2026.3.1)	openai	openai	GPT-5.4	off	32.6%	29	60	3600s	11	1h07m	35.7M	345K	36.1M	$34.79
2026-03-09 05:53	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Opus 4.6	off	71.9%	64	25	3600s	5	1h06m	77.6M	1.5M	79.2M	$98.01
2026-03-09 05:33	OpenClaw (2026.3.1)	openai	openai	GPT-5.4	off	31.5%	28	61	3600s	7	1h04m	34.4M	326K	34.8M	$30.62
2026-03-09 04:27	OpenClaw (2026.3.1)	openai	openai	GPT-5.4	off	30.3%	27	62	3600s	7	1h06m	42.8M	322K	43.1M	$40.66
2026-03-09 03:21	OpenClaw (2026.3.1)	openai	openai	GPT-5.4	off	29.2%	26	63	3600s	5	1h05m	29.6M	333K	30.0M	$30.03
2026-03-09 03:07	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Opus 4.6	off	68.5%	61	28	3600s	4	1h05m	78.4M	1.4M	79.8M	$96.11
2026-03-08 17:46	Terminus-2 (2.0.0)	wandb	moonshotai	Kimi K2.5 (int4) [W&B]	-	48.3%	43	46	3600s	12	1h29m	99.5M	1.7M	101.2M	$17.81
2026-03-08 16:04	Terminus-2 (2.0.0)	wandb	moonshotai	Kimi K2.5 (int4) [W&B]	-	46.1%	41	48	3600s	14	1h41m	120.9M	1.7M	122.6M	$20.20
2026-03-08 14:26	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Opus 4.6	off	75.3%	67	22	3600s	3	1h03m	79.4M	1.4M	80.8M	$94.74
2026-03-08 12:24	Terminus-2 (2.0.0)	wandb	moonshotai	Kimi K2.5 (int4) [W&B]	-	46.1%	41	48	3600s	15	1h26m	101.6M	1.7M	103.3M	$18.00
2026-03-08 12:12	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Sonnet 4.6	-	61.8%	55	34	3600s	10	1h07m	132.2M	2.2M	134.4M	$89.78
2026-03-08 10:25	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Sonnet 4.6	-	59.6%	53	36	3600s	7	1h46m	110.3M	1.9M	112.2M	$76.69
2026-03-08 09:18	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Sonnet 4.6	-	62.9%	56	33	3600s	13	1h06m	98.7M	2.0M	100.7M	$76.46
2026-03-08 08:10	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Sonnet 4.6	-	62.9%	56	33	3600s	6	1h08m	96.7M	1.9M	98.6M	$70.90
2026-03-08 07:14	OpenClaw (2026.3.1)	wandb	moonshotai	Kimi K2.5 (int4) [W&B]	-	34.8%	31	58	3600s	7	1h16m	228.8M	1.6M	230.3M	$37.56
2026-03-08 06:21	Terminus-2 (2.0.0)	anthropic	anthropic	Claude Sonnet 4.6	-	64.0%	57	32	3600s	8	1h48m	77.5M	1.9M	79.4M	$64.31
2026-03-08 05:38	OpenClaw (2026.3.1)	wandb	moonshotai	Kimi K2.5 (int4) [W&B]	-	44.9%	40	49	3600s	11	1h35m	176.6M	1.4M	178.0M	$29.95
2026-03-08 05:12	OpenClaw (2026.3.1)	anthropic	anthropic	Claude Opus 4.6	medium	57.3%	51	38	3600s	10	1h09m	109.9M	1.4M	111.3M	$159.65
2026-03-08 04:03	OpenClaw (2026.3.1)	anthropic	anthropic	Claude Opus 4.6	medium	57.3%	51	38	3600s	7	1h08m	85.2M	1.3M	86.5M	$112.05
2026-03-08 02:53	OpenClaw (2026.3.1)	anthropic	anthropic	Claude Opus 4.6	medium	56.2%	50	39	3600s	10	1h10m	80.8M	1.3M	82.1M	$111.63
2026-03-08 01:44	OpenClaw (2026.3.1)	anthropic	anthropic	Claude Opus 4.6	medium	58.4%	52	37	3600s	8	1h08m	81.1M	1.3M	82.4M	$114.43
2026-03-08 00:37	OpenClaw (2026.3.1)	anthropic	anthropic	Claude Opus 4.6	medium	58.4%	52	37	3600s	5	1h07m	91.8M	1.3M	93.1M	$125.77
2026-03-07 23:25	Terminus-2 (2.0.0)	openai	openai	GPT-5.4	off	43.8%	39	50	3600s	14	1h08m	346.1M	905K	347.0M	$156.79
2026-03-07 22:16	Terminus-2 (2.0.0)	openai	openai	GPT-5.4	off	42.7%	38	51	3600s	12	1h08m	362.6M	878K	363.5M	$144.82
2026-03-07 22:07	OpenClaw (2026.3.1)	anthropic	anthropic	Claude Sonnet 4.6	-	55.1%	49	40	3600s	5	1h20m	97.7M	2.0M	99.7M	$96.83
2026-03-07 20:58	OpenClaw (2026.3.1)	anthropic	anthropic	Claude Sonnet 4.6	-	56.2%	50	39	3600s	3	1h09m	90.6M	2.0M	92.6M	$97.98
2026-03-07 19:47	OpenClaw (2026.3.1)	anthropic	anthropic	Claude Sonnet 4.6	-	51.7%	46	43	3600s	2	1h10m	77.1M	2.0M	79.0M	$71.64
2026-03-07 18:15	OpenClaw (2026.3.1)	anthropic	anthropic	Claude Sonnet 4.6	-	48.3%	43	46	3600s	6	1h31m	126.9M	2.3M	129.2M	$113.18
2026-03-07 15:47	Claude Code (2.1.63)	anthropic	anthropic	Claude Opus 4.6	high	62.9%	56	33	3600s	4	1h05m	103.5M	1.6M	105.2M	$125.04
2026-03-07 13:18	Claude Code (2.1.63)	anthropic	anthropic	Claude Opus 4.6	high	59.6%	53	36	3600s	7	1h09m	108.4M	1.2M	109.5M	$163.56
2026-03-07 11:53	Claude Code (2.1.63)	anthropic	anthropic	Claude Opus 4.6	high	67.4%	60	29	3600s	5	1h24m	109.9M	1.4M	111.4M	$122.24
2026-03-07 10:13	Claude Code (2.1.63)	anthropic	anthropic	Claude Sonnet 4.6	-	53.9%	48	41	3600s	12	1h39m	107.0M	2.1M	109.2M	$84.97
2026-03-07 09:08	Claude Code (2.1.63)	anthropic	anthropic	Claude Sonnet 4.6	-	57.3%	51	38	3600s	4	1h05m	87.8M	1.7M	89.5M	$66.73
2026-03-07 08:02	Claude Code (2.1.63)	anthropic	anthropic	Claude Sonnet 4.6	-	62.9%	56	33	3600s	3	1h05m	97.7M	1.8M	99.4M	$72.46
2026-03-07 06:56	Claude Code (2.1.63)	anthropic	anthropic	Claude Sonnet 4.6	-	57.3%	51	38	3600s	6	1h05m	109.8M	2.2M	112.0M	$81.83

About WolfBench

by Wolfram Ravenwolf – who evaluates models for breakfast, builds agents at night, and preaches AI usefulness all day long.

Welcome to WolfBench – we’re just getting started. What you see here is an early preview with only a handful of models and agents tested so far. We’re continuously expanding the lineup, running fresh evals, and sharing interesting findings and insights along the way. Watch this space.

AI agents are becoming essential tools. Every week, a new model comes out and claims to be “the best at coding” or “SOTA on agentic tasks.” But what does that actually mean for you – the person who’s going to throw real work at these things?

A single score tells you almost nothing.

Most benchmarks give you one number: “Model X scored 42% on Benchmark Y.” Great. But can you rely on it? Was that a lucky run? Would it score the same tomorrow? What’s the floor – the tasks it always nails? What’s the ceiling – what it could do if the stars align?

WolfBench exists because we got tired of meaningless leaderboards. We wanted to know which model, which agent, and which settings actually deliver the best results on real agentic tasks – not just on paper, but in practice, consistently, across multiple runs.

What is it?

WolfBench is an evaluation framework built on top of Terminal-Bench 2.0, a popular agentic benchmark consisting of 89 diverse real-world tasks. These aren’t just coding puzzles. They span the kind of work you’d actually ask an AI agent to do:

System administration: headless terminal interaction, Git server configuration, Nginx request logging
DevOps & infrastructure: package distribution search, database WAL recovery, PyPI server setup
Security: code vulnerability fixes, 7z hash cracking, ELF binary extraction, Git leak recovery
Data & ML ops: financial document processing, HuggingFace model inference, scientific stack modernization
Problem solving: constraint scheduling, adaptive rejection sampling, concurrent task cancellation

The key word is agentic: these tasks require the model to plan, execute shell commands, inspect results, debug failures, and iterate – just like a human developer or sysadmin would. No multiple-choice shortcuts. No toy puzzles. Real work in real sandboxed environments.

Why WolfBench is different

Five-metric framework: Instead of a single average score, we report five complementary metrics that together paint a far more complete picture of what an AI agent can actually do – from the worst-case floor to the theoretical ceiling.
Uniform conditions: Instead of Terminal-Bench 2.0’s default task-specific timeouts and varying sandbox resources, every task in a run gets the same timeout and identical sandbox resources. This ensures scores reflect model and agent capability – not whether an inference endpoint was temporarily overloaded or a sandbox ran out of memory.
Multi-agent comparison: Same model, different agents. Same agent, different models. Different timeouts, concurrency levels, thinking modes. The goal is to understand what matters – not just what scored highest in one particular instance.
Multi-run methodology: A single run is statistically meaningless – variance can swing results widely. We run multiple replicates per configuration to get stable, trustworthy numbers.
Transparency: Every run is collected, classified, and curated with full metadata: tokens consumed, cache hit rates, duration, timeout, concurrency, agent version, thinking mode, etc. Nothing is hidden.

The Five-Metric Framework

Performance is a distribution, not a point. One number can’t capture what an AI agent is truly capable of. Five numbers get a lot closer.

★ Ceiling: What’s theoretically possible?

The union of all tasks ever solved across all runs. If the model solved task A in run 3 and task B in run 5 (but never both in the same run), both count toward the ceiling.

It tells you the theoretical maximum performance this model is capable of with a given agent – even if no single run achieves it. It reveals variance-limited tasks: solvable, but not reliably.

▲ Best-of: What’s the peak in a single run?

The highest score from any individual run.

This is the “marketing number” – but with context. The closer the best-of is to the average, the more consistent the model performs. A large gap between best-of and average means you’re rolling dice every time you run it.

∅ Average: What can you normally expect?

The mean score across all valid runs.

This is the most commonly reported metric – and it is useful, but only with enough runs to be stable. With a single run? It’s a coin flip.

▼ Worst-of: How bad can a single run get?

The lowest score from any individual run.

This is the opposite of best-of – the floor, the worst case. The gap between worst-of and best-of defines the full score range across all runs. A narrow range means predictable performance; a wide range means you’re rolling dice.

■ Solid: What does it always get right?

Tasks that the model solves across all runs – the rock-solid base with zero variance.

The higher the solid base, the more dependable the agent is. These are the tasks you can confidently delegate and expect success every time. A model with a high solid base and moderate average is often more reliable in practice than one with a high average but low solid base – because you know what you’re getting.

WolfIndex

WolfIndex is a dataset-independent, optional ranking lens over the same five metrics, not a replacement for them. Toggle it above the chart to sort by 50% Average + 20% Solid + 5% Worst-of + 5% Best-of + 20% Ceiling. Average remains the anchor; Solid and Ceiling preserve task-level reliability and breadth; Best and Worst retain a small signal for whole-run outliers.

The index is computed from the unrounded binary task results and only rounded to one decimal place for display. Official ranking requires exactly five complete runs from one configuration; other run counts remain provisional, accumulating, or nonstandard, while incomplete or mixed legacy groups show N/A. Formula: wolfindex-balanced-1.0.0.

Reading the Chart

The five metrics are shown for each model/configuration as stacked bar segments from the rock-solid base up to the ceiling. Optional 3D modes add token volume, run cost, total runtime, or tokens plus cost as depth: token mode splits opaque non-cached input, translucent cache-read input, and output tokens behind a visible input/output separator, cost mode uses the total cost for depth, time mode uses total runtime for depth, and the combined mode adds a neutral gray cost shadow behind the token-depth bar. The spread between the segments tells you as much as the numbers themselves:

Tight spread (metrics close together) = consistent, predictable AI agent
Wide spread (big gap between solid and ceiling) = high variance, unreliable
High ceiling, low average = the model can do it, but usually doesn’t – needs more runs or better settings
High solid, close to average = rock-solid workhorse you can count on

The Bottom Line

Performance is more complex than a single average score – and the decisions you make based on benchmarks deserve better data than that. WolfBench gives you five angles on every model and configuration, so you can form a more complete and realistic judgement of what an AI agent will actually deliver when you put it to work.

Because at the end of the day, you don’t just want to know which model scored the highest. You want to know which one you can trust.

What’s Next

We will continuously add models and agents to the chart, publish the traces and evals on W&B Weave, and release regular blog posts detailing interesting and insightful findings.

This benchmark offers enormous potential for discovery. For instance: Why does xhigh reasoning improve GPT 5.4’s performance while max effort degrades Opus 4.6’s results? How does Claude Code fare when running a GPT or Gemini model compared to running directly with Opus or Sonnet – or Codex with Claude or Gemini? Is a “cheap” model actually cost-effective if it consumes far more tokens than a more expensive alternative? How does quantization affect performance of local models in agentic tasks?

So many possibilities for analysis – and for posting about it! Stay tuned – and if you want to be the first to know when new results come in, follow me on X and LinkedIn.

Inference and sandbox compute sponsored by CoreWeave: The Essential Cloud for AI.
Additional sandbox compute by Daytona – Secure Infrastructure for Running AI-Generated Code.
Built with Harbor for orchestration, Terminal-Bench 2.0 for tasks, and W&B Weave for tracking.
Charts and dashboards generated with marimo notebooks.
Explore the complete data and tooling suite on our WolfBench GitHub.

WolfBench (2026-09-10)