AI Model Changelog

GPT-6 Sol released with a 1,050,000-token context window, 128K max output, and an April 20, 2026 knowledge cutoff

API pricing set at $2.00 per 1M input tokens and $10.00 per 1M output tokens, half of GPT-5.6 Sol; cached input $0.20 per 1M tokens; OpenAI says the cut is permanent rather than promotional

GPT-6 Luna released with a 1,050,000-token context window, 128K max output, and a May 18, 2026 knowledge cutoff; supports reasoning effort from none to max

API pricing set at $0.10 per 1M input tokens and $0.50 per 1M output tokens, half of GPT-5.6 Luna; cached input $0.01 per 1M tokens

Succeeded by GPT-6 Sol, which launched at half the price

Succeeded by GPT-6 Luna, which launched at half the price

Claude Opus 5.5 released with a 1M context window, up to 128K output tokens, always-on adaptive thinking, and a June 2026 knowledge cutoff

Pricing set at $4.00 per 1M input tokens and $20.00 per 1M output tokens, 20% below Opus 5; cache reads drop to $0.20 per 1M tokens

Breaking changes from Opus 5: thinking can no longer be disabled, forced tool use returns an error, thinking blocks are bound to the model and conversation, and the computer_20251124 tool was removed; default effort changed from high to medium

Anthropic announced Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks

Moved to the legacy model list after the release of Claude Opus 5.5; remains available for compatibility and migration comparisons

Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS released to general availability with voice design, voice replication, and a Voices endpoint

Access to Gemini 2.5 models limited to users who have actively used them before; not deprecated and still served through the API

Access to Gemini 2.5 models limited to users who have actively used them before; not deprecated and still served through the API

Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking audio-to-audio models released for the Live API

DeepSeek V4.1 Flash released with a 1M context window, up to 384K output, native vision, and support for both thinking and non-thinking modes; call it via the deepseek-flash model name

API pricing reduced to $0.30 per 1M input tokens and $1.20 per 1M output tokens at peak hours ($0.15 and $0.60 off-peak); cache hits $0.006 per 1M tokens

Planned routing of V4 Pro requests to V4.1 Flash after September 14, 2026 was reversed; V4 Pro remains on the API with unchanged billing

V4 Flash and V4 Flash Vision Exp retired on September 10, 2026; the deepseek-v4-flash and deepseek-v4-flash-vision-exp names are temporarily routed to V4.1 Flash

GPT-6 Astra rolled out to paid ChatGPT users in a restricted version that declines certain cybersecurity prompts

GPT-6 Astra unveiled and released as a limited preview with a 1,050,000-token context window, 128K max output, and an April 30, 2026 knowledge cutoff; OpenAI says it meets the Critical cybersecurity threshold under its Preparedness Framework

API pricing set at $10.00 per 1M input tokens and $50.00 per 1M output tokens (2.5x GPT-5.6 Sol); cached input $1.00, Batch at half price, Fast mode at 2x; requests above about 272K input tokens billed at $20.00 input and $75.00 output

Gemini 3.8 Flash released to general availability in the Gemini API as the recommended model for new projects

API pricing set at $0.75 per 1M input tokens and $3.75 per 1M output tokens through December 31, 2026, rising to $1.50 and $7.50 from January 1, 2027

Claude Fable 5.1 released alongside the restricted-access Claude Mythos 5.1, with a 1M context window, 128K max output, always-on adaptive thinking, and a June 2026 knowledge cutoff

Pricing unchanged from Fable 5 at $10.00 per 1M input tokens and $50.00 per 1M output tokens; cache reads cut 75% from $1.00 to $0.25 per 1M tokens

Forced tool use behavior changed (breaking change); Anthropic recommends starting workloads on Opus and escalating to Fable 5.1

Introductory rate of $2.00 input and $10.00 output per 1M tokens made the standard price; the previously scheduled increase to $3.00 and $15.00 was dropped

Superseded by Claude Fable 5.1 and moved to the legacy model list

Claude Mythos 5.1 released as the trusted-access sibling of Fable 5.1 for vetted cyberdefense and life-science organizations