| # Changelog |
|
|
| All notable changes to the LLM Model Comparison dataset will be documented in this file. |
|
|
| Format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/). |
|
|
| ## [Q1 2026 v2] - 2026-02-18 |
|
|
| ### Changed |
|
|
| - Major update: replaced all models with latest generations |
| - Added GPT-4.1 and GPT-4.1 mini (1M context window) |
| - Added o3 and o4-mini reasoning models |
| - Added Claude 4.5 family (Sonnet, Haiku, Opus) |
| - Added Gemini 2.5 Pro and Gemini 2.5 Flash |
| - Added Llama 4 Scout and Llama 4 Maverick (10M context window) |
| - Added DeepSeek V3 (MIT, open source) |
| - Added Mistral Large 3 (Apache 2.0, open source) |
| - Added Command A (Cohere) |
|
|
| ### Removed |
|
|
| - Removed deprecated models: GPT-4o, Claude 3.5 family, Gemini 2.0/1.5, Llama 3.x, older Mistral and Cohere models |
|
|
| ## [Q1 2026] - 2026-02-15 |
|
|
| ### Added |
|
|
| - Initial release with 16 LLM models across 7 providers |
| - 22 fields per model including benchmark scores, latency metrics, and API feature flags |
| - Providers: OpenAI, Anthropic, Google, Meta, DeepSeek, Mistral AI, Cohere |
|
|
| --- |
|
|
| Published by [Salt Technologies AI](https://www.salttechno.ai) |
|
|
