BlogLLM Benchmark Scores as Published, Side by Side
Harvey's Legal Agent Benchmark: Results as Published
Every result the catalog holds for Harvey's Legal Agent Benchmark: 30 models from 10 developers, read from 15 pages. Each number is as printed by the company named beside it; nothing is ranked or rescored. Models are listed by developer, newest first.
Area Knowledge WorkMaintainer HarveyIntroduced 2026Data version 2026103001, as of 2026-10-30
Harvey's Legal Agent Benchmark tests AI agents on long, multi-step legal work. Each task gives the agent a short instruction and a client matter with case files, and asks for a work product that a reviewing lawyer would check, across more than 1,200 tasks in 24 legal practice areas. Expert-written rubrics grade each deliverable, and a task counts as passed only when every criterion passes.
Harvey's Legal Agent Benchmark
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPT-6 AstraOpenAI, released 2026-09-03 | 5.4% | Google, Gemini 4 Argonchecked 2026-10-29; printed on 1 more page | - |
| GPT-5.6 SolOpenAI, released 2026-07-09 | 2.5% | Anthropic, Claude Opus 5checked 2026-10-15; printed on 4 more pages | - |
| GPT-5.6 TerraOpenAI, released 2026-07-09 | 0.8% | Google, Gemini 3.8 Flash model cardchecked 2026-10-15; printed on 1 more page | - |
| GPT-5.5OpenAI, released 2026-04-23 | 2.1% | Anthropic, Claude Fable 5 and Claude Mythos 5checked 2026-10-15 | - |
| Claude Opus 5.5Anthropic, released 2026-09-22 | 3.8% | Google, Gemini 4 Argonchecked 2026-10-29 | - |
| Claude Fable 5.1Anthropic, released 2026-09-01 | 6.7% | Google, Gemini 4 Argonchecked 2026-10-29; printed on 2 more pages | - |
| Claude Opus 5Anthropic, released 2026-07-24 | 11.7% | Anthropic, Claude Opus 5checked 2026-10-15; printed on 1 more page | - |
| Claude Sonnet 5Anthropic, released 2026-06-30 | 5% | Google, Gemini 3.8 Flash model cardchecked 2026-10-15; printed on 1 more page | - |
| Claude Fable 5Anthropic, released 2026-06-09 | 13.3% | Anthropic, Claude Opus 5checked 2026-10-15; printed on 2 more pages | - |
| Claude Mythos 5Anthropic, released 2026-06-09 | 13.3% | Anthropic, Claude Fable 5 and Claude Mythos 5checked 2026-10-15 | - |
| Claude Opus 4.8Anthropic, released 2026-05-28 | 10.4% | Anthropic, Claude Fable 5 and Claude Mythos 5checked 2026-10-15; printed on 1 more page | - |
| Gemini 4 ArgonGoogle, released 2026-09-30 | 19.6% | Google, Gemini 4 Argonchecked 2026-10-29 | - |
| Gemini 3.8 FlashGoogle, released 2026-09-02 | 10% | Google, Gemini 3.8 Flash model cardchecked 2026-10-15 | - |
| Gemini 3.7 FlashGoogle, released 2026-08-13 | 8.8% | Google, Gemini 3.8 Flash model cardchecked 2026-10-15 | - |
| Gemini 3.1 Pro PreviewGoogle, released 2026-02-19 | 0% | Anthropic, Claude Fable 5 and Claude Mythos 5checked 2026-10-15 | - |
| Grok 4.7xAI, released 2026-09-21 | 19.6% | xAI, Grok 4.7checked 2026-10-25; printed on 1 more page | Effort: xhigh |
| Grok 4.6xAI, released 2026-08-12 | 15.8% | xAI, Introducing Grok 4.6checked 2026-10-15; printed on 2 more pages | Effort: high |
| Grok 4.5xAI, released 2026-07-16 | 12.9% | xAI, Introducing Grok 4.6checked 2026-10-15 | Effort: high |
| DeepSeek-V4-Pro-0813DeepSeek | 7.5% | Mistral AI, Mistral Large 4 (preview)checked 2026-10-15 | - |
| Qwen3.8-MaxAlibaba, released 2026-08-03 | 10.4% | Mistral AI, Mistral Large 4 (preview)checked 2026-10-15 | - |
| Kimi K3Moonshot AI, released 2026-07-16 | 12.9% | Mistral AI, Mistral Large 4 (preview)checked 2026-10-15 | - |
| GLM-5.3Z.ai, released 2026-08-18 | 8.3% | Mistral AI, Mistral Large 4 (preview)checked 2026-10-15 | - |
| GLM-5.2Z.ai, released 2026-06-16 | 7.1% | Mistral AI, Mistral Large 4 (preview)checked 2026-10-15 | - |
| Mistral Large 4 (preview)Mistral AI, released 2026-10-06 | 15.8% | Mistral AI, Mistral Large 4 (preview)checked 2026-10-15 | - |
Harvey's Legal Agent Benchmark (criterion pass rate)
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPT-5.6 SolOpenAI, released 2026-07-09 | 87.2% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: max |
| GPT-5.6 TerraOpenAI, released 2026-07-09 | 85.2% | Google, Gemini 3.7 Flash model cardchecked 2026-10-15 | - |
| GPT-5.5OpenAI, released 2026-04-23 | 86.3% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: xhigh |
| Claude Sonnet 5.5Anthropic, released 2026-09-28 | 92.1% | Anthropic, Claude Sonnet 5.5 System Cardchecked 2026-10-19 | Effort: high; Harness: Artificial Analysis harness |
| 93.1% | Anthropic, Claude Sonnet 5.5 System Cardchecked 2026-10-19 | Effort: max; Harness: Artificial Analysis harness | |
| Claude Opus 5.5Anthropic, released 2026-09-22 | 91.2% | Anthropic, Claude Opus 5.5 System Cardchecked 2026-10-25 | Effort: max; Dataset revision: held-out set |
| Claude Fable 5.1Anthropic, released 2026-09-01 | 90.81% | Anthropic, System Card: Claude Fable 5.1 & Claude Mythos 5.1checked 2026-10-25 | Dataset revision: public set |
| 93.3% | Anthropic, System Card: Claude Fable 5.1 & Claude Mythos 5.1checked 2026-10-25 | Effort: xhigh; Dataset revision: held-out set | |
| Claude Sonnet 5Anthropic, released 2026-06-30 | 90.1% | Google, Gemini 3.7 Flash model cardchecked 2026-10-15 | - |
| Claude Fable 5Anthropic, released 2026-06-09 | 93.6% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: max; Fallback: w/ fallback |
| Claude Opus 4.8Anthropic, released 2026-05-28 | 91.1% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: max |
| Gemini 3.7 FlashGoogle, released 2026-08-13 | 90.7% | Google, Gemini 3.7 Flash model cardchecked 2026-10-15 | - |
| Gemini 3.6 FlashGoogle, released 2026-07-21 | 85.1% | Google, Gemini 3.7 Flash model cardchecked 2026-10-15 | - |
| Kimi K3Moonshot AI, released 2026-07-16 | 94.6% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: max |
| GLM-5.2Z.ai, released 2026-06-16 | 91% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: max |
Harvey LAB-AA v1.1 (tasks meeting every criterion)
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPT-6 AstraOpenAI, released 2026-09-03 | 8.6% | Meta, Meta Muse Wonder model cardchecked 2026-10-19 | Effort: max |
| Claude Sonnet 5.5Anthropic, released 2026-09-28 | 10% | Anthropic, Claude Sonnet 5.5 System Cardchecked 2026-10-19 | Effort: max; Harness: Artificial Analysis harness |
| 11.7% | Anthropic, Claude Sonnet 5.5 System Cardchecked 2026-10-19 | Effort: high; Harness: Artificial Analysis harness | |
| Claude Opus 5.5Anthropic, released 2026-09-22 | 8.3% | Anthropic, Claude Opus 5.5 System Cardchecked 2026-10-25; printed on 1 more page | Effort: max |
| Claude Fable 5.1Anthropic, released 2026-09-01 | 16.7% | Anthropic, System Card: Claude Fable 5.1 & Claude Mythos 5.1checked 2026-10-25; printed on 1 more page | Effort: xhigh |
| Muse WonderMeta, released 2026-10-12 | 10% | Meta, Meta Muse Wonder model cardchecked 2026-10-19 | Effort: max |
| Muse Wonder FullMeta, released 2026-10-12 | 13.1% | Meta, Meta Muse Wonder model cardchecked 2026-10-19 | Effort: max |
Harvey Legal Agent (public set, all-pass rate)
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| Claude Fable 5.1Anthropic, released 2026-09-01 | 19.09% | Anthropic, System Card: Claude Fable 5.1 & Claude Mythos 5.1checked 2026-10-25; printed on 1 more page | - |
| DeepSeek-V4.1-Flash-1017DeepSeek, released 2026-10-17 | 21% | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Effort: max; Harness: Stirrup |
| Kimi K3.1 ProMoonshot AI, released 2026-10-23 | 23.7% | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Effort: max; Harness: Stirrup |
Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.