Changelog

What changed in each calibration — keep us accountable

Expanded to 25 models + price history tracking

  • Expanded coverage from 16 to 25 models, adding Gemini 3 Flash family, Grok 4.1 Fast, Qwen3-7 Max, GLM-5.1/5.2, and Kimi K2.5/K2.6.
  • Introduced price-history fields: DeepSeek V4 Pro and GPT-5.4 now show their price-drop timeline (e.g. DeepSeek permanently cut to 1/4 of original).
  • Added "starting price" column and region legend to the Channels page.

Full refresh to 2026 generation

  • Updated prices for all 16 models across official + aggregator channels (OpenAI, Anthropic, Google, xAI, DeepSeek, Alibaba, Zhipu, ByteDance, Moonshot, OpenRouter).
  • Added benchmark sub-scores (coding/math/reasoning/vision) for GPT-5.5, GPT-5.4, Claude Opus 4.8, Claude Sonnet 4.6.
  • Added rate-limit / TTFT / throughput specs for OpenAI and Anthropic frontier models.

New models added

  • Added GPT-5.5 and Claude Opus 4.8 (2026 frontier releases).
  • Refreshed Artificial Analysis Intelligence Index scores.

Launch — cost calculator & plan comparison

  • Launched ModelMarket with per-token cost calculator and subscription-plan comparison — capabilities no major competitor offers.
  • Established 16 data sources (official pricing pages + OpenRouter/LMSYS/Artificial Analysis), every figure traceable to a dated source.