AI Models

Claude 5.5 family and GPT-6 models: pricing, context windows and head-to-head comparisons.

Last updated 2026-10-02 · Vendor data sourced from Anthropic and OpenAI

Claude 5.5 Family: Fable, Opus & Sonnet Compared

The Claude 5.5 generation spans three models released in September 2026, all sharing a 1M-token context window and built-in extended thinking. Pick a tier based on your quality, speed and budget requirements.

Fable 5.1Opus 5.5Sonnet 5.5
Input / output price$10 / $50 per MTok$4 / $20 per MTok$2 / $10 per MTok
Context window1M tokens1M tokens1M tokens
Max output128K tokens128K tokens (300K on Batch API)128K tokens (300K on Batch API)
SpeedSlowerModerateFast
ThinkingAlways onAlways onAdaptive
Default efforthighmediumhigh
Best forDemanding reasoningAgentic codingDay-to-day coding

See Fable 5.1 vs Opus 5.5 and Opus 5.5 vs Sonnet 5.5 and Sonnet 5 vs 5.5 for benchmark breakdowns and per-task recommendations.

Anthropic Models

Claude Fable 5.1

For demanding reasoning and long-horizon agentic work

Input / output
$10 / $50 per MTok
Cache read
$0.25 per MTok
Context
1M tokens
Released
Sep 1, 2026
Pricing, benchmarks & cost calculator →

Claude Opus 5.5

For long-running agentic coding and knowledge work

Input / output
$4 / $20 per MTok
Cache read
$0.20 per MTok
Context
1M tokens
Released
Sep 22, 2026
Pricing, benchmarks & cost calculator →

Claude Sonnet 5.5

The best combination of speed and intelligence

Input / output
$2 / $10 per MTok
Cache read
$0.20 per MTok
Context
1M tokens
Released
Sep 28, 2026
Pricing, benchmarks & cost calculator →

Claude Sonnet 5

Anthropic's balanced reasoning model (succeeded by Sonnet 5.5)

Input / output
$2 / $10 per MTok
Cache read
$0.20 per MTok
Context
1M tokens
Released
Jun 30, 2026

OpenAI Models

GPT-6 Astra

OpenAI's most capable model for complex reasoning and end-to-end work

Input / output
$10 / $50 per MTok
Cache read
$1 per MTok
Context
1.05M tokens
Released
Sep 3, 2026
Pricing, benchmarks & cost calculator →

GPT-6 Luna

OpenAI's lightweight reasoning model — 20x cheaper than GPT-6 Sol

Input / output
$0.10 / $0.50 per MTok
Cache read
$0.01 per MTok
Context
1.05M tokens
Released
Sep 22, 2026
Pricing, benchmarks & cost calculator →

GPT-6 Sol

OpenAI's efficient reasoning model (superseded by GPT-6.1 Sol)

Input / output
$2 / $10 per MTok
Cache read
$0.20 per MTok
Context
1.05M tokens
Released
Sep 22, 2026
Pricing, benchmarks & cost calculator →

GPT-6.1 Sol

OpenAI's latest efficient reasoning model, successor to GPT-6 Sol

Input / output
$2 / $10 per MTok
Cache read
$0.10 per MTok
Context
1.05M tokens
Released
Sep 29, 2026
Pricing, benchmarks & cost calculator →

Compare Two Models Side by Side

Want a verdict instead of a list? The AI model comparison hub has head-to-head pages with benchmarks, per-task cost and a recommendation for each use case.

AI Models FAQ

  • What is Claude 5.5?

    Claude 5.5 is the current generation of Anthropic's Claude models, released in September 2026. The family includes three tiers: Fable 5.1 ($10/$50 per MTok) for demanding reasoning, Opus 5.5 ($4/$20) for agentic coding and knowledge work, and Sonnet 5.5 ($2/$10) for fast, cost-efficient everyday tasks. All three share a 1M-token context window and 128K max output.

  • Which Claude 5.5 model should I use?

    Start with Sonnet 5.5 — it's the fastest and cheapest, and handles most coding, analysis and data tasks well. Move to Opus 5.5 when you need higher benchmark scores or long-running agentic workflows. Reserve Fable 5.1 for tasks that specifically need deep, sustained reasoning chains, as it costs 2.5× more than Opus and is slower.

  • How much does Claude 5.5 cost?

    Sonnet 5.5 costs $2/$10 per million tokens, Opus 5.5 costs $4/$20, and Fable 5.1 costs $10/$50. All three offer 50% batch discounts and prompt caching. Cache reads cost $0.20–$0.25 per million tokens across the family.

  • What is the difference between this page and the model comparison page?

    This page is a directory: one entry per model with its price, context window and links. The model comparison page puts two models side by side with benchmarks and a recommendation for each use case.

  • Where do the prices and specs come from?

    From each vendor's official documentation. Every value on a model page is tied to a source URL and a last-verified date, and benchmark and speed figures come from Artificial Analysis or the vendor's own announcement, labeled as such.

  • Why do some models have a full page and others only comparison links?

    Full model pages are added one at a time. Until a model has its own page, the directory links to every comparison that covers it.