claude-cli · fast
CLAUDE-CLI

claude-haiku

Anthropic Haiku (subscription)

⚖ COMPARE MODELS
VALS INDEX
54.3% +/- 21.5
Accuracy
VALS INDEX
40.4s
Latency (median)
$VALS INDEX
$22.4783
Cost / Test (1P-est)
SCORE STATUSscored
TEST SETS12
RECORDED TRIALS1112
CUSTOM PASS74%

Industry benchmarks

Accuracyattempts · responded
GPQA_DIAMOND GPQA Diamond
36.7%
30 attempts · 29 responded
HUMANEVAL_PLUS HumanEval+
48.0%
100 attempts · 50 responded
IFEVAL IFEval
78.3%
30 attempts · 30 responded
Industry benchmark Bars = accuracy %, black cap = that score

Custom 9-task suite

Mean /5n=trials
CUSTOM json_strict
5.00
115 trials
CUSTOM reasoning_multistep
4.99
100 trials
CUSTOM code_gen_long
4.76
119 trials
CUSTOM agentic_tool_use
4.72
85 trials
CUSTOM summarize
4.66
116 trials
CUSTOM refactor_existing_code
3.95
95 trials
CUSTOM agentic_prompt
3.74
95 trials
CUSTOM code_debug
1.78
125 trials
CUSTOM creative_write
1.38
102 trials
Custom suite · scored 0–5 by consensus (m3 + Fable-5) ★ = task scored a 5 at least once · sorted by mean score

Cross-judge validation

m3 vs Fable-5 Δ

No judge bias detected on this model — m3 and Fable-5 agree across all tasks.

Per-task distribution

Score

All trials on this page were run through the harness. Industry suites use deterministic family-specific reference graders (no LLM judge). The custom suite is graded by consensus of minimax-m3 + claude-fable-5. See /industry/ for the full benchmark suite.