claude-cli · long-context
CLAUDE-CLI

claude-fable

Anthropic Fable 5 (subscription, 1M context — for long-form craft)

⚖ COMPARE MODELS
VALS INDEX
73.7% +/- 25.1
Accuracy
VALS INDEX
18.1s
Latency (median)
$VALS INDEX
$9.6823
Cost / Test (1P-est)
SCORE STATUSscored
TEST SETS12
RECORDED TRIALS852
CUSTOM PASS88%

Industry benchmarks

Accuracyattempts · responded
GPQA_DIAMOND GPQA Diamond
96.7%
30 attempts · 30 responded
HUMANEVAL_PLUS HumanEval+
47.0%
100 attempts · 50 responded
IFEVAL IFEval
77.4%
62 attempts · 62 responded
Industry benchmark Bars = accuracy %, black cap = that score

Custom 9-task suite

Mean /5n=trials
CUSTOM json_strict
5.00
65 trials
CUSTOM reasoning_multistep
5.00
45 trials
CUSTOM agentic_prompt
4.99
75 trials
CUSTOM code_debug
4.82
105 trials
CUSTOM agentic_tool_use
4.74
50 trials
CUSTOM summarize
4.73
75 trials
CUSTOM code_gen_long
4.71
95 trials
CUSTOM refactor_existing_code
3.99
70 trials
CUSTOM creative_write
2.29
80 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.