Alpheva Legal: Legal Reasoning Benchmark

Legal reasoning across evolving matters.

Can a model revise a grounded legal work product as the record evolves?

Claude Opus 5 Max(Anthropic)

62.1%

Claude Fable 5.1 Max(Anthropic)

60.8%

Claude Fable 5(Anthropic)

55.7%

Kimi K3 Max(Moonshot AI)

54.1%

Grok 4.6 High(xAI)

53.8%

GPT-5.6 Sol Max(OpenAI)

51.1%

Muse Spark 1.3 xHigh(Meta)

50.3%

Gemini 3.8 Flash High(Google)

47.9%

Grok 4.5 High(xAI)

43.2%

Muse Spark 1.1 xHigh(Meta)

42.3%

Claude Sonnet 5(Anthropic)

40.9%

Claude Opus 4.8(Anthropic)

38.2%

Claude Opus 4.7(Anthropic)

35.8%

GPT-5.5(OpenAI)

35.1%

Gemini 3.1 Pro(Google)

21.9%

Inkling Medium(Thinking Machines)

21.9%

Mean

Overview

Alpheva Legal evaluates how models identify controlling issues, apply authorities to the record, revise conclusions as facts and arguments change, and ground legal work in text and visual evidence across multi-stage matters.

Evaluation focus

The benchmark is organized around the capabilities that determine whether a model can complete the full workflow reliably.

01Issue, rule, application, and conclusion
02Revision across multi-stage matters
03Grounding in legal authorities and records
04Reasoning over visual exhibits

Protocol

Protocol: Alpheva v0.1
Materials: Authorities + case files
Scope: Multi-stage matters

Model results

Model	Provider	Mean
Claude Opus 5 Max	Anthropic	62.1%
Claude Fable 5.1 Max	Anthropic	60.8%
Claude Fable 5	Anthropic	55.7%
Kimi K3 Max	Moonshot AI	54.1%
Grok 4.6 High	xAI	53.8%
GPT-5.6 Sol Max	OpenAI	51.1%
Muse Spark 1.3 xHigh	Meta	50.3%
Gemini 3.8 Flash High	Google	47.9%
Grok 4.5 High	xAI	43.2%
Muse Spark 1.1 xHigh	Meta	42.3%
Claude Sonnet 5	Anthropic	40.9%
Claude Opus 4.8	Anthropic	38.2%
Claude Opus 4.7	Anthropic	35.8%
GPT-5.5	OpenAI	35.1%
Gemini 3.1 Pro	Google	21.9%
Inkling Medium	Thinking Machines	21.9%

Results and methodology

Scores are reported against the protocol and versioned environment defined above. Compare results within this benchmark and metric only.