Evidence and rankings

Compare models with the right evidence.

Inspect measured coding runs, provider-specific usage, or reported features and token-price preferences. Each view answers a different question and keeps its source and limits visible.

Metadata heuristics · v0.7 / explorer v0.6

Historical coding runs

Published Aider Polyglot results with test dates and configurations.

Provider request share

Observed use within a named gateway, with coverage and history.

Feature and price preferences

A reproducible metadata heuristic; adjust the weights below.

Feature and price preference explorer

value reasoning agents open weights cheapest context window

RankModelIndexPriceContextMovement

GLM-5.3-FlashZhipu AI

98.1

$0.0163 / $0.1821OpenRouter · USD / 1M1M—

Qwen3.6 27B FP8Alibaba

97.2

$0 / $0InferX · USD / 1M262K—

Gemma 4 31B IT FP8Google

97.2

$0 / $0InferX · USD / 1M262K—

Gemma 4 26B A4B ITGoogle

96.9

$0.024 / $0.10Kilo Gateway · USD / 1M262K—

Qwen3.5 4BAlibaba

96.8

$0.04 / $0.07EmpirioLabs AI · USD / 1M262K—

Qwen3.5 9BAlibaba

96.6

$0.05 / $0.15NanoGPT · USD / 1M262K—

Qwen3.6 35B-A3BAlibaba

96.4

$0.045 / $0.30engy · USD / 1M262K—

Qwen3.8 27BAlibaba

96.4

$0.045 / $0.32engy · USD / 1M262K—

GLM 5.3 Flash UncensoredZhipu AI

96.3

$0.20 / $0.80NanoGPT · USD / 1M1.05M—

Qwen3.8 Flash NextAlibaba

96.3

$0.10 / $0.20Vultr · USD / 1M262K—

Gemma 4 31B ITGoogle

96.2

$0.10 / $0.25LLM Gateway · USD / 1M262K—

Qwen3.5 35B-A3BAlibaba

96.2

$0.0564 / $0.4512AIHubMix · USD / 1M262K—

Gemma 4 E2B ITGoogle

96.1

$0.04 / $0.08Amazon Bedrock · USD / 1M131K—

Qwen3 VL 30B A3B ThinkingAlibaba

95.9

$0.1046 / $0.4185SiliconFlow (China) · USD / 1M262K—

Qwen3.5 27BAlibaba

95.7

$0.0846 / $0.6768AIHubMix · USD / 1M262K—

Qwen3.5 122B-A10BAlibaba

95.3

$0.1126 / $0.9008AIHubMix · USD / 1M262K—

Qwen3.5 397B-A17BAlibaba

94.9

$0.1644 / $0.9864AIHubMix · USD / 1M262K—

Qwen3.6 35B A3B FP8Alibaba

94.8

$0.17 / $1.10NEAR AI Cloud · USD / 1M262K—

Qwen3.6 27BAlibaba

94.6

$0.15 / $1.35DevPass (LLM Gateway) · USD / 1M262K—

Qwen/Qwen3-VL-32B-ThinkingAlibaba

94.4

$0.1494 / $1.4945SiliconFlow (China) · USD / 1M262K—

Kimi K2.6Moonshot AI

94.3

$0.275 / $1.10routing.run · USD / 1M262K—

Kimi K2.7 CodeMoonshot AI

94.3

$0.275 / $1.10routing.run · USD / 1M262K—

Muse Glimmer 30BMeta

94.3

$0.20 / $0.80EmpirioLabs AI · USD / 1M131K—

Qwen3 Omni 30B A3B ThinkingAlibaba

93.2

$0.25 / $0.97Novita AI · USD / 1M66K—

Command A PlusCohere

93.1

$0.30 / $1.50OpenRouter · USD / 1M128K—

Kimi K2.5Moonshot AI

93.0

$0.375 / $2NanoGPT · USD / 1M262K—

Magistral Small 1.2Mistral AI

92.3

$0.50 / $1.50Amazon Bedrock · USD / 1M128K—

Qwen3 VL 235B A22B ThinkingAlibaba

91.9

$0.287 / $2.867Merge Gateway · USD / 1M131K—

Qwen 3.8 27B UncensoredAlibaba

91.9

$1 / $3NanoGPT · USD / 1M1.05M—

Qwen 3.8 27B Uncensored ThinkingAlibaba

91.9

$1 / $3NanoGPT · USD / 1M1.05M—

How this score is calculated

Capability coverage, logarithmic price efficiency, context size, and openness are normalized to a 0–100 scale. Capability coverage counts four reported feature flags; context and openness each contribute only through their own weight. Missing prices are unranked when price has a positive weight. These preferences do not measure task quality.