Favur Evals
favur.dev

Agent role · weight 3/100 · rubric v15

📐 Spec

Decision-artifact producer — concrete values only (no TBD/hedging), with a self-check step.

What it's scored on

Output quality
Quality of the decision artifact it wrote and deliverables.
Efficiency
Token volume, output per request, cost, context efficiency, request-error and hallucination rates.

Model leaderboard

Every model ranked by Spec score
#ModelScore / 10CoveragenLast seen
1Meta Muse Spark 1.17.8196%32026-07-19
2Xiaomi Mimo v2.57.3196%22026-07-26
3Google Gemini Flash 3.57.1176%12026-07-08
4DeepSeek Deepseek Flash v47.0179%42026-07-22
5Moonshot AI Kimi Code k2.76.9276%12026-07-08
6xAI Grok 4.56.9091%42026-07-22
7Google Gemini Flash 3 (preview)6.8672%42026-07-09
8OpenAI Gpt Terra 5.66.7276%12026-07-12
9DeepSeek Deepseek Pro v46.6076%12026-07-05
10Google Gemini Flash 3.66.3196%42026-07-24
11Anthropic Claude Sonnet 56.1686%42026-07-23
12Google Gemini Flash Lite 3.56.0996%12026-07-21
13Google Gemini Pro 3.1 (preview)5.8168%12026-06-28
14Google Gemini Flash Lite 3.15.1263%22026-07-15

† low confidence — sparse role or low data coverage (renormalized, never zero-filled).

Per-run history (newest first)

Every appearance of the Spec role, newest first
DateRunSoWModelScore / 10
2026-07-26solarSystem__multi-xiaomi-mimo-v2.5-7.26.2026solarSystemXiaomi Mimo v2.57.17
2026-07-24circles__xiaomi-multi-model-2.5-2.5-pro-7.24.2026circlesXiaomi Mimo v2.57.45
2026-07-24circles__google-gemini-3.6-flash-7.23.2026circlesGoogle Gemini Flash 3.66.61
2026-07-23solarSystem/multi-anthropic-sonnet-gemini-3.5-flash-lite-7.22.2026solarSystemAnthropic Claude Sonnet 55.87
2026-07-22circles__anthropic-claude-sonnet-5-7.21.2026circlesAnthropic Claude Sonnet 56.38
2026-07-22circles__x-ai-grok-4.5-7.22.2026circlesxAI Grok 4.56.88
2026-07-22circles__deepseek-deepseek-v4-flash-7.22.2026circlesDeepSeek Deepseek Flash v47.28
2026-07-22the2048__gemini-multi-model-7-22-2026the2048Google Gemini Flash 3.65.92
2026-07-21solarSystem__google-multi-model-7.21.2026solarSystemGoogle Gemini Flash 3.66.42
2026-07-21circles__x-ai-grok-4.5-7.21.2026circlesGoogle Gemini Flash 3.66.29
2026-07-21circles__google-gemini-3.5-flash-lite-7.21.2026circlesGoogle Gemini Flash Lite 3.56.09
2026-07-19circles__x-ai-grok-4.5-7.18.2026circlesxAI Grok 4.56.89
2026-07-19the2048__meta-muse-spark-1.1-7.19.2026the2048Meta Muse Spark 1.17.85
2026-07-17circles__x-ai-grok-4.5-7.16.2026circlesxAI Grok 4.57.19
2026-07-17solarSystem__meta-muse-spark-1.1-7.17.2026solarSystemMeta Muse Spark 1.17.89
2026-07-17circles__meta-muse-spark-1.1-7.16.2026circlesMeta Muse Spark 1.17.69
2026-07-15circles__google-multi-model-7.16.2026circlesGoogle Gemini Flash Lite 3.13.61
2026-07-13solarSystem__google-gemini-3.1-flash-lite-7.12.2026solarSystemGoogle Gemini Flash Lite 3.16.64
2026-07-12circles__anthropic-claude-sonnet-5-7.11.2026circlesAnthropic Claude Sonnet 56.28
2026-07-12circles__openai-gpt-5.6-terra-7.11.2026circlesOpenAI Gpt Terra 5.66.72
2026-07-10circles__x-ai-grok-4.5-7.10.2026circlesxAI Grok 4.56.65
2026-07-10circles__deepseek-deepseek-v4-flash-7.10.2026circlesDeepSeek Deepseek Flash v47.43
2026-07-09circles__google-gemini-3-flash-preview-7.8.2026circlesGoogle Gemini Flash 3 (preview)6.89
2026-07-08circles__google-gemini-3.5-flash-7.8.2026circlesGoogle Gemini Flash 3.57.11
2026-07-08circles__anthropic-claude-sonnet-5-7.8.2026circlesAnthropic Claude Sonnet 56.12
2026-07-08circles__moonshotai-kimi-k2.7-code-7.8.2026circlesMoonshot AI Kimi Code k2.76.92
2026-07-05circles__deepseek-deepseek-v4-pro-7.5.2026circlesDeepSeek Deepseek Pro v46.60
2026-07-05circles__google-gemini-3-flash-preview-7.5.2026circlesGoogle Gemini Flash 3 (preview)7.15
2026-07-04circles__deepseek-deepseek-v4-flash-7.3.2026circlesDeepSeek Deepseek Flash v46.55
2026-06-28circles__google-gemini-3.1-pro-preview-6.28.2026circlesGoogle Gemini Pro 3.1 (preview)5.81
2026-06-28solarSystem__google-gemini-3-flash-preview-6.27.2026solarSystemGoogle Gemini Flash 3 (preview)6.27
2026-06-26circles__google-gemini-3-flash-preview-6.26.2026circlesGoogle Gemini Flash 3 (preview)7.14
2026-06-26the2048__deepseek-deepseek-v4-flash-6.26.2026the2048DeepSeek Deepseek Flash v46.79

Every run on Favur Evals is scored by the same deterministic engine, on the same Statements of Work. The benchmark is self-funded — no vendor sponsorship, credits, or grants.