★VALS INDEX
43.0%
+/- 36.2
Accuracy
⏱VALS INDEX
5.5s
Latency (median)
$VALS INDEX
$0.0305
Cost / Test (1P-est)
SCORE STATUSscored
TEST SETS13
RECORDED TRIALS683
CUSTOM PASS40%
Industry benchmarks
Accuracyattempts · responded
GPQA_DIAMOND
GPQA Diamond
17.2%
180 attempts · 115 responded
HUMANEVAL_PLUS
HumanEval+
90.0%
40 attempts · 36 responded
IFEVAL
IFEval
52.6%
180 attempts · 140 responded
MMLU_PRO
MMLU-Pro
12.0%
200 attempts · 135 responded
Custom 9-task suite
Mean /5n=trials
CUSTOM
json_strict
★
4.09
11 trials
CUSTOM
code_debug
★
3.91
11 trials
CUSTOM
refactor_existing_code
-
3.67
3 trials
CUSTOM
agentic_prompt
★
3.18
11 trials
CUSTOM
agentic_tool_use
-
3.00
3 trials
CUSTOM
code_gen_long
★
1.36
11 trials
CUSTOM
reasoning_multistep
★
1.36
11 trials
CUSTOM
summarize
★
1.36
11 trials
CUSTOM
creative_write
-
0.00
11 trials
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.