skynet · general
SKYNET

ollama-cl/kimi-k2.5

Kimi k2.5 (Moonshot)

⚖ COMPARE MODELS
VALS INDEX
47.6% +/- 21.6
Accuracy
VALS INDEX
19.8s
Latency (median)
$VALS INDEX
$0.5073
Cost / Test (1P-est)
SCORE STATUSscored
TEST SETS13
RECORDED TRIALS424
CUSTOM PASS42%

Industry benchmarks

Accuracyattempts · responded
GPQA_DIAMOND GPQA Diamond
23.1%
52 attempts · 34 responded
HUMANEVAL_PLUS HumanEval+
36.1%
36 attempts · 13 responded
IFEVAL IFEval
68.2%
151 attempts · 146 responded
MMLU_PRO MMLU-Pro
63.0%
100 attempts · 100 responded
Industry benchmark Bars = accuracy %, black cap = that score

Custom 9-task suite

Mean /5n=trials
CUSTOM reasoning_multistep
5.00
11 trials
CUSTOM summarize
3.64
11 trials
CUSTOM code_gen_long
3.45
11 trials
CUSTOM code_debug
3.42
12 trials
CUSTOM json_strict
3.08
12 trials
CUSTOM agentic_tool_use -
0.67
3 trials
CUSTOM refactor_existing_code -
0.67
3 trials
CUSTOM agentic_prompt -
0.64
11 trials
CUSTOM creative_write -
0.45
11 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.