claude-cli · reasoning
CLAUDE-CLI

claude-opus-5

Anthropic Opus 5 via Claude Max (low-effort launch wave; 1M API context)

⚖ COMPARE MODELS
VALS INDEX
80.6% +/- 7.6
Accuracy
VALS INDEX
15.9s
Latency (median)
$VALS INDEX
$1.3818
Cost / Test (1P-est)
SCORE STATUSscored
TEST SETS13
RECORDED TRIALS1493
CUSTOM PASS89%

Industry benchmarks

Accuracyattempts · responded
GPQA_DIAMOND GPQA Diamond
84.8%
198 attempts · 198 responded
HUMANEVAL_PLUS HumanEval+
78.0%
164 attempts · 164 responded
IFEVAL IFEval
71.3%
541 attempts · 541 responded
MMLU_PRO MMLU-Pro
88.4%
500 attempts · 500 responded
Industry benchmark Bars = accuracy %, black cap = that score

Custom 9-task suite

Mean /5n=trials
CUSTOM agentic_prompt
5.00
10 trials
CUSTOM code_gen_long
5.00
10 trials
CUSTOM json_strict
5.00
10 trials
CUSTOM reasoning_multistep
5.00
10 trials
CUSTOM agentic_tool_use
4.90
10 trials
CUSTOM summarize
4.90
10 trials
CUSTOM code_debug
4.70
10 trials
CUSTOM refactor_existing_code -
3.00
10 trials
CUSTOM creative_write
2.00
10 trials
Custom suite · scored 0–5 by consensus (m3 + Fable-5) ★ = task scored a 5 at least once · sorted by mean score

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.