Skip to content
Model Pulse

Weekly recap · auto-written from the diff log

This week in AI pricing

957 changes hit the catalog this week: 283 repricings, 316 launches, 8 sunsets, and 260 spec shifts. The busiest day was 2026-09-13 with 329 moves.

Full changelogSubscribe via RSS

Cut of the week

DeepSeek V4 Flash 0731: $0.22 → $0.035 /M input (−84%)

All its provider prices →

The labs moved

first-party pricing

  1. 1.DeepSeek V4 Flash$0.14→$0.15+7%
  2. 2.DeepSeek V4 Flash Vision Exp$0.14→$0.15+7%

The street moved

gateway markup

  1. 1.Solar Pro 4$0.03→$0.09+200%
  2. 2.Kimi K2.6$0.22→$0.6+173%
  3. 3.Qwen3 Coder Next$0.22→$0.5+127%
  4. 4.DeepSeek V3.2$0.28→$0.62+121%
  5. 5.Qwen3-Coder 480B-A35B Instruct$0.22→$0.45+105%
  6. 6.GLM-5.3-Flash$0.075→$0.15+100%
  7. 7.Gemma 3 27B IT$0.12→$0.23+92%
  8. 8.Gemma 3 12B IT$0.05→$0.09+80%

Launches 316

new modelDeepSeek V4.1 Flash (Consensus Protocol)via llmgateway-providers ·
new modelGemma 3 4Bvia merge-gateway ·
new modelKimi K3 (Tencent Cloud)via llmgateway-providers ·
new modelGemma 3 27B IT (Amazon Bedrock)via edenai ·
new modelGLM-5V-Turbovia llmgateway ·
new modelGLM-5.3via agentrouter ·
new modelVoxtral Mini 3B 2507 (Amazon Bedrock, US)via edenai ·
new modelGemma 3 4B ITvia deepinfra ·
new modelinclusionAI: Ling 3.0 Flash VLvia kilo ·

Sunsets 8

  • deprecatedGLM-4.7-Flashvia deepinfra
  • deprecatedGLM-5via deepinfra
  • deprecatedQwen3.8 Max Previewvia alibaba-token-plan-cn
  • deprecatedOmen Alphavia opencode-go
  • deprecatedQwen3.8 Max Previewvia alibaba-token-plan
  • deprecatedMiniMax-M2.7via deepinfra
  • deprecatedGPT-5.1 Chatvia zenmux
  • deprecatedKimi K2.5via deepinfra

Spec shifts

contextGLM-5.2via kilo ·
  • context: 203K → 1.05M tokens
  • output: 182K → 131K tokens
contextNemotron 3 Ultra 550B A55Bvia kilo ·
  • context: 256K → 203K tokens
  • output: 33K → 183K tokens
contextDeepSeek V4 Pro 0813via kilo ·
  • context: 1.05M → 1.02M tokens
  • output: 393K → 384K tokens
contextQwen: Qwen3 14Bvia kilo ·
  • context: 131K → 41K tokens
  • output: 8K → 16K tokens
contextGLM-5.3via kilo ·
  • context: 1.05M → 1.05M tokens
  • output: 131K → 944K tokens
capabilityNova 2 Lite (JP)via amazon-bedrock ·
  • last_updated: 2025-12-02 → 2025-12-01

Last week’s tape

967 changes last week (219 repricings, 176 launches, 2 sunsets).

  • repricedGLM-4.62026-09-07
  • capabilityClaude Fable Latest (Claude Fable 5.1)2026-09-07
  • capabilityDeepSeek V4 Flash 0731 (FlexAI)2026-09-07
  • capabilityNemotron 3 Ultra 550B A55B (Nebius)2026-09-07
  • contextQwen3.5 397B-A17B2026-09-07
+962 more in the changelog →