Agent Upstart

Cool things people are building with agents — and practical help getting the most out of them.

Coding playbooks · Reference table

Model prices: what the major models cost

Written November 5, 2026

Input, output, and cached prices for the models coding agents actually use, each row linked to the vendor's own pricing page.

All prices are in US dollars per million tokens. Every row links to the vendor page the price came from. Prices move often and some are introductory, so treat this table as a starting point and open the vendor page before budgeting a workload.

How to read this table

Cached input is cheaper than fresh input everywhere, but the rules differ per vendor: what counts as cached, how long a cache entry lives, and whether cache writes cost extra all vary. A model that looks cheap on input price can cost more than a dearer one once caching and output volume are counted. Introductory prices end: Gemini 4 Argon doubles on December 1, and the LongCat and Hy4 rows below are discounted rates, not list prices. The table shows prices, not a ranking — context windows, rate limits, and routing differ too much for a cheapest-model claim to mean anything.

The table

ModelInput / 1MCached input / 1MOutput / 1MNotes
GPT-6.1 Sol$2.00$0.10$10.00OpenAI API pricing
GPT-6.1 Astra$10.00$1.00$50.00OpenAI API pricing
Claude Sonnet 5.5$2.00$0.10$10.00Anthropic API pricing
Claude Opus 5.5$4.00$0.20$20.00Anthropic API pricing
Gemini 4 Argon$2.00$0.10$10.00$4.00 / $0.20 / $20.00 from December 1, 2026. Google Gemini API pricing
Kimi K3.1 Pro$2.00$0.20$10.001,048,576-token context. Kimi K3.1 Pro pricing
Kimi K3.1 Pro Highspeed$4.00$0.20$20.001,048,576-token context. Cache writes $2.00 (5 min) / $4.00 (1 hour). Kimi K3.1 Pro pricing
Mistral Large 4$0.70$0.07$2.101M-token context. Model id mistral-large-4. Mistral Large 4 announcement
LongCat-2.5$0.30$0.006$1.20Discounted price, limited time, no end date given; prices subject to change. 1M-token context. LongCat API pay-as-you-go pricing
Ripple-65B$0.20$0.05$0.60Preview, sold through OpenRouter; single provider. 1,048,576-token context. Ripple-65B on OpenRouter
Hy4$0.834$0.042$2.501List price; currently shown at 50% off on OpenRouter ($0.417 / $0.021 / $1.251). 1,048,576-token context. Hy4 on OpenRouter

What this table does not do

It does not include the cost of the agent around the model: the harness, the tools, the retries, and the context re-reads that a long session accumulates. For the full picture, see what an agent costs and the agent cost estimator. To estimate a specific workload from the rates above, use the token cost calculator. If you pay through a subscription instead of the API, see bring your own Anthropic key and the GPT-6.1 Sol ultrafast Pro tiers. For the Argon price change in full, see Gemini 4 Argon price doubling on December 1.

Prices here are snapshots. When a row's price changes on the vendor page, the row changes with it — dates that have passed are removed as the list is updated.

Related tools

Sources

Back to: Coding playbooks hub · Also: Token cost calculator · Agent cost estimator · What an agent costs