AI Model
Cost Matrix
A practical reference for understanding AI model pricing, capabilities, and when to use which model. All prices per 1M tokens (USD). Current as of August 27, 2026.
Created by Sean Bair
A quick reference for students and friends working out AI model costs.
Questions? Get in touch
Questions? Get in touch
22 Models
3 Providers
$0.10 Cheapest Input
$180 Most Expensive Output
Pricing Matrix
All prices per 1M tokens (USD). Click a model string to copy it.
| Provider | Model | Model String | Input $/1M | Output $/1M | Context | Speed | Intelligence | Best For | Notes |
|---|---|---|---|---|---|---|---|---|---|
| Anthropic | Haiku 4.5 | claude-haiku-4-5 | $1 | $5 | 200K | Fast classification, routing, simple extraction, high-volume preprocessing | Cheapest Claude. Great for bulk preprocessing. | ||
| Anthropic | Sonnet 5NEW | claude-sonnet-5 | $2 | $10 | 1M | Production workhorse: coding, agents, analysis, document processing | Near-Opus quality. $2/$10 is now the standard price; the planned Sep 2026 increase was cancelled. | ||
| Anthropic | Opus 4.7 | claude-opus-4-7 | $5 | $25 | 1M | Complex reasoning, long-horizon agents, research, vision | Previous-gen flagship. Adaptive thinking, xhigh effort. | ||
| Anthropic | Opus 4.8 | claude-opus-4-8 | $5 | $25 | 1M | Autonomous agents, knowledge work, memory | Fast mode + mid-session system prompts. | ||
| Anthropic | Opus 5NEW | claude-opus-5 | $5 | $25 | 1M | Current Opus flagship: complex reasoning, autonomous agents, research | Same price as Opus 4.8. Fast mode available. | ||
| Anthropic | Fable 5NEW | claude-fable-5 | $10 | $50 | 1M | Most demanding reasoning, overnight autonomous runs, hardest problems | Anthropic’s most capable model. Thinking always on. 30-day retention. | ||
| OpenAI | GPT-5.4 nano | gpt-5.4-nano | $0.2 | $1.25 | 400K | Bulk processing, simple transforms, labeling at scale | Cheapest OpenAI. Cached input $0.02. | ||
| OpenAI | GPT-5.4 mini | gpt-5.4-mini | $0.75 | $4.5 | 400K | Mid-tier tasks, moderate reasoning, structured output | Good balance for OpenAI lightweight work. | ||
| OpenAI | GPT-5.3 Codex | gpt-5.3-codex | $1.75 | $14 | 400K | Coding agents, code generation, repository work | Tuned for code. Cached input $0.175. | ||
| OpenAI | GPT-5.4 | gpt-5.4 | $2.5 | $15 | 400K | Mid-tier flagship: analysis, coding, agentic workflows | Cached input $0.25. Strong all-rounder. | ||
| OpenAI | GPT-5.5 | gpt-5.5 | $5 | $30 | 1M | Frontier reasoning, coding, agentic workflows | Cached input $0.50. Tiered above 272K tokens. | ||
| OpenAI | GPT-5.5 Pro | gpt-5.5-pro | $30 | $180 | 1M | Hardest problems, deepest reasoning, no cost ceiling | Most expensive. Only when nothing else suffices. | ||
| OpenAI | GPT-5.6 lunaNEW | gpt-5.6-luna | $0.20 / 0.40 | $1.20 / 1.80 | 1M | Cost-sensitive workloads at 5.6-generation quality | Cached input $0.02. Tiered above 272K tokens. | ||
| OpenAI | GPT-5.6 terraNEW | gpt-5.6-terra | $2.00 / 4.00 | $12.00 / 18.00 | 1M | OpenAI mainline: balanced intelligence and cost for production work | Cached input $0.20. Tiered above 272K tokens. | ||
| OpenAI | GPT-5.6 solNEW | gpt-5.6-sol | $4.00 / 8.00 | $20.00 / 30.00 | 1M | OpenAI flagship: complex professional work, deep reasoning, agents | Promo $4/$20 through Nov 21, 2026. Tiered above 272K tokens. | ||
| Gemini 2.5 Flash-Lite | gemini-2.5-flash-lite | $0.1 | $0.4 | 1M | Bulk classification, translation, summarization, routing | Cheapest model anywhere. 1M context. | |||
| Gemini 3.1 Flash-LiteNEW | gemini-3.1-flash-lite | $0.25 | $1.5 | 1M | Cheap agentic tasks, classification with newer reasoning | Latest Flash-Lite. 1M context. | |||
| Gemini 2.5 Flash | gemini-2.5-flash | $0.3 | $2.5 | 1M | Document analysis, data extraction, high-volume with thinking | GA stable. Great price/perf. 1M context. | |||
| Gemini 3 Flash | gemini-3-flash-preview | $0.5 | $3 | 1M | Best-value reasoning: coding, agents, multimodal | Preview. Pro-level benchmarks at Flash price. | |||
| Gemini 3.5 FlashNEW | gemini-3.5-flash | $1.5 | $9 | 1M | Latest Flash: fast agents, coding, multimodal, high-volume | Released May 2026. Newest Flash tier. | |||
| Gemini 2.5 Pro | gemini-2.5-pro | $1.25 / 2.50 | $10.00 / 15.00 | 1M | Deep reasoning, complex coding, long-context analysis | Tiered: doubles >200K tokens. | |||
| Gemini 3.1 ProNEW | gemini-3.1-pro-preview | $2.00 / 4.00 | $12.00 / 18.00 | 1M | Most capable Gemini: complex reasoning, ARC-AGI-2 leader, agents | Preview. Tiered: doubles >200K tokens. Same price as 3 Pro. |
Model Selection Decision Guide
Which model should you use for which task? A practical framework.
Simple classification / routing / tagging
Claude Haiku 4.5$1 / $5
Fast, cheap, accurate for bounded outputs
Bulk data labeling / high-volume extraction
Gemini 2.5 Flash-Lite or GPT-5.4 nano$0.10–0.20 in
Cheapest tokens anywhere. 1M context on Gemini.
JSON extraction from text / email
Claude Haiku 4.5$1 / $5
Structured output at lowest Claude cost
Production RAG answers (student-facing)
Claude Sonnet 5$2 / $10
Near-Opus quality at the standard Sonnet price
Long-form content (blogs, reports)
Claude Sonnet 5$2 / $10
Best writing quality at the Sonnet price point
Email / outreach drafting
Claude Sonnet 5$2 / $10
Personality and nuance without flagship cost
Document analysis + summarization
Gemini 3 Flash or Claude Sonnet 5$0.50–2 in
Gemini 3 Flash is the value pick; Sonnet for accuracy
Agentic multi-tool / coding workflows
Claude Sonnet 5 or GPT-5.4$2–2.50 in
Best tool-use and long-horizon focus for the price
Dedicated coding agent
GPT-5.3 Codex or Claude Opus 5$1.75–5 in
Codex is tuned for code; Opus 5 for hardest refactors
Very long context (>200K tokens)
Gemini 3.1 Pro or GPT-5.5$2–5 in
1M windows. Gemini tiers double past 200K but stay cheap.
Complex reasoning / research
Claude Opus 5 or Gemini 3.1 Pro$2–5 in
Opus 5 most autonomous; Gemini 3.1 Pro cheaper
Hardest problems / overnight autonomous runs
Claude Fable 5$10 / $50
Most capable model. Give it your unsolved problems.
Frontier reasoning on OpenAI
GPT-5.6 sol or GPT-5.5 Pro$4–30 in
Sol is the new flagship; Pro for max capability, no cost limit.
Best value for coding + agents
Gemini 3 Flash$0.50 / $3.00
Pro-level benchmarks at Flash price
High-volume with thinking (budget)
Gemini 2.5 Flash$0.30 / $2.50
Cheap reasoning model, 1M context, GA stable