Overview
30+
Models Tracked
5
Providers
6
Routing Dimensions
15
Decision Rules
Data Gaps Detected (auto-audit vs capability-data.md)
Terminal-Bench chartDS V4 Flash listed twice (duplicate entry)Removed duplicate
Terminal-Bench chart10 models missing (no TB score): Kimi K2.7, Grok Build, MiMo v2.5, Ring 2.6 1T, ERNIE 5.1, Llama 4 Maverick, Qwen3-Coder-Next, Haiku 4.5, DS V4 Pro Max, Agnes 2.0Added as '?' entries
Terminal-Bench chartMistral Large 3 (TB 12.0) missing from chartAdded
SWE-Pro chartSonnet 5 (63.2) missing from chart despite being in pricing tableAdded
SWE-Pro chartHaiku 4.5 (39.5) and DS V4 Flash (52.6) missing from chartAdded
MCP-Atlas chartOpus 4.8 showed 82.2 but capability-data.md says 77.8Corrected to 77.8
MCP-Atlas chartDS V4 Flash (69.0) and Mistral Large 3 (70.4) missing from chartAdded
GPT-5.5 TB scoreChart showed 83.4, capability-data.md says 84.0Corrected to 84.0
Pricing tableFable 5 listed as 'SUSPENDED' — it was restored July 1Changed to 'Classifier gap'
Pricing tableGrok Build 0.1 appeared twice (duplicate row)Removed duplicate
Pricing tableMissing models: Haiku 4.5, DS V4 Pro Max, Fugu Ultra, Cohere North Mini Code, Mythos 5, Grok 4.20Added
Provider panelsFable 5 listed as 'Suspended' in ZenMux and Nous panelsChanged to 'Restored'
Provider panelsMissing Sonnet 5, Step 3.5 Flash, Haiku 4.5 from provider listsAdded where available
Pricing tiersSonnet 5 missing from premium tier, Step 3.5 Flash missing from budget tierAdded
Composer 2.5Listed in pricing table with colspan layout — not a standalone API modelMoved to notes on Grok Build row