No price changes.
15 models · real prices · updated daily
What each AI model costs, and what it's actually good at.
Real prices per million tokens, context limits, and concrete use cases for 15 models and 8 coding agents. No benchmark talk.
Daily digest
What changed since the last update.
Auto-generated on every nightly run. Last update September 13, 2026.
None in the last day. Latest: DeepSeek V4 Flash Vision Exp (2026-09-10).
Picks unchanged.
Quick answers
The picks, analyzed nightly.
AI reads every price and spec in the catalog and refreshes these picks on each run. Last analyzed September 13, 2026.
Claude Sonnet 5
$2 in · $10 out
Offers terminal agents and whole-repo refactors at $2 per 1M input tokens.
GLM-5.3-Flash
$0.075 in · $0.25 out
Has the lowest combined price of $0.075 input and $0.25 output per 1M tokens.
DeepSeek V4 Pro
$0.435 in · $0.87 out
Provides daily coding agent reasoning on a budget for only $0.87 per 1M output tokens.
GPT-6 Astra
$10 in · $50 out
Delivers complex reasoning and research-heavy work with a 1.05M token context.
GPT-5.6 Terra
$2 in · $12 out
Supports 1.05M context tokens for general coding and mixed text at $2 per 1M input.
Gemini 3.8 Flash
$0.75 in · $3.75 out
Features a dedicated free-tier learning use case along with a 1.04M token context.
The catalog
A model for every job.
15 models with real prices per 1M tokens. Every card shows what to use it for and when to skip it.
Claude Sonnet 5
Anthropic · 2026-06-29
Use for
Whole-repo refactors · Features with tests · Terminal agents
Skip for
Bulk batch jobs — output costs $10 per 1M
in$2out$10cached in$0.2
Claude Opus 5
Anthropic · 2026-07-24
Use for
Hard debugging · Architecture calls · Long agent runs
Skip for
Everyday edits — 2.5x Sonnet's output price
in$5out$25cached in$0.5
GPT-6 Astra
OpenAI · 2026-09-04
Use for
Complex reasoning · Research-heavy work
Skip for
Budget work — $50 per 1M output
in$10out$50cached in$1
GPT-5.6 Terra
OpenAI · 2026-07-09
Use for
General coding · Mixed text and code
Skip for
High-volume loops — Luna is 10x cheaper
in$2out$12cached in$0.2
GPT-5.6 Luna
OpenAI · 2026-07-09
Use for
Autocomplete · Draft code · Simple scripts
Skip for
Complex multi-file changes
in$0.2out$1.2cached in$0.02
Gemini 3.8 Flash
Google · 2026-09-02
Use for
Large file processing · Fast iteration · Free-tier learning
Skip for
Top-tier reasoning — use Gemini Pro instead
in$0.75out$3.75cached in$0.075
Grok 4.6
xAI · 2026-08-12
Use for
Chat-style coding help · 500K-context tasks
Skip for
Code-heavy repos — half the context of 1M rivals
in$2out$6cached in$0.5
DeepSeek V4 Pro
DeepSeek · 2026-08-12
Use for
Daily coding agent · Reasoning on a budget · Long files
Skip for
Only when you need the absolute best output
in$0.435out$0.87cached in$0.004
DeepSeek V4 Flash
DeepSeek · 2026-09-10
Use for
High-volume edits · Tests and boilerplate · Cheap agents
Skip for
Tricky logic — weaker than V4 Pro
in$0.15out$0.6cached in$0.003
GLM-5.3
Z.ai · 2026-08-14
Use for
Agentic workflows · Code review · 1M-context tasks
Skip for
Cost-sensitive volume — Flash is 4x cheaper
in$1.4out$4.4cached in$0.26
GLM-5.3-Flash
Z.ai · 2026-08-26
Use for
Bulk edits · Renames and cleanup · Learning projects
Skip for
Complex reasoning
in$0.075out$0.25cached in$0.015
Kimi K3
Moonshot · 2026-07-16
Use for
Frontier work with open weights · 1M-context tasks
Skip for
Budget work — $15 per 1M output
in$3out$15cached in$0.3
Kimi K2.7 Code
Moonshot · 2026-06-12
Use for
Code generation · Refactors · Agent loops
Skip for
Very large files — limited to 262K context
in$0.95out$4cached in$0.19
Qwen3.8 Max
Alibaba · 2026-08-03
Use for
General coding · Multilingual work
Skip for
Cheapest tier — GLM Flash costs a fraction
in$2out$6cached in$0.25
Muse Spark 1.3
Meta · 2026-09-02
Use for
Open-weight experimentation · Cost-controlled scale
Skip for
Frontier-quality needs
in$1.25out$4.25cached in$0.15
Prices updated September 13, 2026 · Source: models.dev. "Use for" and "Skip for" are editorial; prices and limits are live.
Head to head
Compare models side by side.
Pick up to three models. The winner of every row gets a star.
The market
Every model, one table.
The 15 catalog models plus every release from the 9 tracked labs in the last 120 days. Newest first. Verified September 13, 2026.
| Model | Lab | Released | In $/1M | Out $/1M | Context | Card |
|---|---|---|---|---|---|---|
| DeepSeek V4 Flash | DeepSeek | 2026-09-10 | $0.15 | $0.6 | 1M | ★ |
| DeepSeek V4 Flash Vision Exp | DeepSeek | 2026-09-10 | $0.15 | $0.6 | 1M | — |
| DeepSeek V4.1 Flash | DeepSeek | 2026-09-10 | $0.15 | $0.6 | 1M | — |
| GPT-6 Astra | OpenAI | 2026-09-04 | $10 | $50 | 1.05M | ★ |
| Gemini 3.8 Flash | 2026-09-02 | $0.75 | $3.75 | 1.05M | ★ | |
| Muse Spark 1.3 | Meta | 2026-09-02 | $1.25 | $4.25 | 1.05M | ★ |
| Muse Spark 1.3 Contributor | Meta | 2026-09-02 | $0.1 | $0.2 | 1.05M | — |
| Claude Fable 5.1 | Anthropic | 2026-09-01 | $10 | $50 | 1M | — |
| GLM-5.3-Flash | Z.ai | 2026-08-26 | $0.075 | $0.25 | 1M | ★ |
| Qwen3.8 Flash | Alibaba | 2026-08-26 | $0.15 | $0.47 | 1M | — |
| GLM-5.3 | Z.ai | 2026-08-14 | $1.4 | $4.4 | 1M | ★ |
| Gemini 3.7 Flash | 2026-08-13 | $0.75 | $3.75 | 1.05M | — | |
| Gemini Flash Latest | 2026-08-13 | $0.75 | $3.75 | 1.05M | — | |
| Grok 4.6 | xAI | 2026-08-12 | $2 | $6 | 500K | ★ |
| DeepSeek V4 Pro | DeepSeek | 2026-08-12 | $0.435 | $0.87 | 1M | ★ |
| Muse Spark 1.2 | Meta | 2026-08-05 | $1.25 | $4.25 | 1.05M | — |
| Muse Spark 1.2 Contributor | Meta | 2026-08-05 | $0.1 | $0.2 | 1.05M | — |
| Qwen3.8 Max | Alibaba | 2026-08-03 | $2 | $6 | 1M | ★ |
| DeepSeek V4 Flash 0731 | Alibaba | 2026-07-31 | $0.2 | $0.4 | 1M | — |
| Claude Opus 5 | Anthropic | 2026-07-24 | $5 | $25 | 1M | ★ |
| Gemini 3.6 Flash | 2026-07-21 | $0.75 | $3.75 | 1.05M | — | |
| Gemini 3.5 Flash Lite | 2026-07-21 | $0.3 | $2.5 | 1.05M | — | |
| Gemini Flash-Lite Latest | 2026-07-21 | $0.3 | $2.5 | 1.05M | — | |
| Kimi K3 | Moonshot | 2026-07-16 | $3 | $15 | 1.05M | ★ |
| GPT-5.6 Terra | OpenAI | 2026-07-09 | $2 | $12 | 1.05M | ★ |
| GPT-5.6 Luna | OpenAI | 2026-07-09 | $0.2 | $1.2 | 1.05M | ★ |
| Claude Sonnet 5 | Anthropic | 2026-06-29 | $2 | $10 | 1M | ★ |
| Kimi K2.7 Code | Moonshot | 2026-06-12 | $0.95 | $4 | 262K | ★ |
★ marks the models with a full card above. The rest came from the nightly lab scan.
By the numbers15
models tracked
9
labs covered
8
agents compared
$0.325
cheapest 1M in + out (GLM-5.3-Flash)
A 200x gap between the cheapest output (GLM-5.3-Flash at $0.25) and the priciest (GPT-6 Astra at $50).
The agents
The eight agents worth knowing.
The agent is the tool you type into; you pay for the model behind it. Here is what each one costs and runs.
opencode
Any provider
Price
Free, open source
Best for
opencode.ai ↗You bring your own API key and own the setup
Claude Code
Claude only
Price
Included with Claude Pro ($20/mo)
Best for
code.claude.com ↗Running Sonnet 5 or Opus 5 in a terminal
Cursor
Claude, GPT, Gemini
Price
Free tier, Pro $20/mo
Best for
cursor.com ↗Beginners — autocomplete plus agent mode in one editor
Windsurf
Claude, GPT
Price
Free tier, paid from $15/mo
Best for
windsurf.com ↗Fast inline completions with a familiar VS Code feel
OpenAI Codex
GPT-5.6, GPT-6
Price
Free tier with ChatGPT
Best for
github.com/openai/codex ↗Delegating tasks that run in your repo or on GitHub
Gemini CLI
Gemini
Price
Free tier, generous daily quota
Best for
github.com/google-gemini/gemini-cli ↗A real agent at $0 — best free starting point
Cline
Any provider, ClinePass
Price
Free, open source
Best for
cline.bot ↗Flat $9.99/mo open-weight models inside VS Code
GitHub Copilot
GPT, Claude
Price
Free tier, Pro $10/mo
Best for
github.com/features/copilot ↗Autocomplete in the editor you already use
Tip: most agents are free or open source. You pay for the model, not the harness — except ClinePass at a flat $9.99.
Recommended for you
Dial in your setup. Get your stack.
Slide your budget, set your priorities, and the AI picks the best stack for you.
What are you building?
Monthly budget $0
Priority Balanced
Your experience
Make it yours
Pick a theme, change everything.
10 themes restyle the whole site at once. Click one, or press T to flip through them.
Your choice is saved in this browser. Try T right now.
Sources
Where the data comes from.
Prices change weekly. These are the pages we check first.
models.dev ↗Open database of model pricing and limitsOpenRouter ↗Unified catalog and routing for 400+ modelsAnthropic pricing ↗Claude API rates, updated with each releaseOpenAI pricing ↗GPT model rates and batch discountsGoogle AI pricing ↗Gemini API rates and free tierClinePass ↗Flat-rate open-weight subscription