Sign InOpen Brain
VercelEngineering PostOfficial Source

AI Gateway: GPT-5.6 pricing and speed updates

Vercel cut Luna and Terra token prices and raised Sol fast-mode speed without changing model IDs, so existing agent workloads inherit the changes without code edits.

Vercel · Jul 30, 2026
Open Source Open MarkdownOpen JSON
Source Summary

AI Gateway cut **GPT-5.6 Luna pricing by 80%** and **GPT-5.6 Terra pricing by 20%** across short and long contexts. Luna short-context rates are $0.20 input and $1.20 output per million tokens; Terra is $2 and $12.

Practical Implication

Recheck model routing and budgets for agent workloads: Luna and Terra may now fit tasks previously assigned elsewhere. Existing requests receive the new rates because model IDs did not change.

Agent-Ready Context
AI Gateway cut **GPT-5.6 Luna pricing by 80%** and **GPT-5.6 Terra pricing by 20%** across short and long contexts. Luna short-context rates are $0.20 input and $1.20 output per million tokens; Terra is $2 and $12.

Recheck model routing and budgets for agent workloads: Luna and Terra may now fit tasks previously assigned elsewhere. Existing requests receive the new rates because model IDs did not change.

**GPT-5.6 Sol fast mode is now 2.5x**, up from 1.5x, but its price is unchanged. The material provides only short-context dollar figures, so consult the full pricing detail before estimating long-context spend.
Connected Context · Feed7 Judgment

This materially changes the routing economics of existing GPT-5.6 workloads without requiring model-ID migrations: Luna becomes far cheaper, Terra moderately cheaper, and Sol fast mode faster at unchanged price. It justifies rerunning task-level cost and latency evaluations, but the missing long-context figures and outcome-quality evidence prevent price changes alone from determining the new routing policy.

AI Gateway adds unified fast mode supportUnified fast mode supplies the common request mechanism through which Sol’s increased fast-mode multiplier can affect existing coding-agent latency choices.Notion's Token Town — Sarah Sachs, NotionThe price cuts strengthen the case for revisiting routes, while Token Town cautions that per-token savings alone do not establish lower end-to-end agent cost.TRACE-ROUTER: Task-Consistent and Adaptive Online Routing for Agentic AITRACE-Router provides an outcome-based method for testing whether the newly cheaper Luna or Terra should own an entire task instead of routing from price alone.Routing rules now available on AI GatewayRouting rules make it possible to apply a reassessed Luna, Terra, or Sol policy centrally without changing application code, matching the operational consequence of the new economics.
Context Map
infracoding#gateways#model-selection
Uncertainty
**GPT-5.6 Sol fast mode is now 2.5x**, up from 1.5x, but its price is unchanged. The material provides only short-context dollar figures, so consult the full pricing detail before estimating long-context spend.