Standard Compute
← All models

GPT-5.6 Luna

Budget

The fast, cheap tier of the GPT-5.6 family — routine work and subagents at $1/$6 per million.

OpenAI
Maker
85/100
Overall
59
Value score
$2.50
Blended /1M
1.1M
Context
Proprietary
Weights
Pricing: $1.00 in · $6.00 out / 1M tokens · synced from OpenRouter · updated 2026-07-17

Scorecard

Output Quality8.4/10 · 74% community
Agentic Ability8.3/10 · 74% community
Speed9.3/10 · 79% community
Value for $9.0/10 · 77% community
Reliability9.1/10 · 78% community

GPT-5.6 Luna is the speed-and-cost tier of OpenAI's July 2026 family: summarization, drafting, routing, and the many small calls agents make. At $1/$6 per million tokens with the family's large context, it competes directly with Haiku 4.5 and the open-weight value models for high-volume agent steps, with an optional pro reasoning mode for harder moments.

Strengths
  • Very low price for the capability ($1/$6 per M)
  • Fast responses suited to high-volume loops
  • Large context unusual at this price tier
Trade-offs
  • Clear quality step below Terra and Sol
  • Very new — behavior under long agent chains still settling

Agents that run Luna well

Cheap, fast subtasks, routing, and subagents in OpenAI-based stacks.

Best model for CodexBest model for ClineBest model for OpenCodeBest model for HermesBest model for OpenClaw

GPT-5.6 Luna FAQ

Is GPT-5.6 Luna good for AI agents?
Cheap, fast subtasks, routing, and subagents in OpenAI-based stacks. On our editorial scoring it rates 8.3/10 for agentic ability and 8.4/10 for output quality. Very low price for the capability ($1/$6 per M).
How much does GPT-5.6 Luna cost?
Via OpenRouter, GPT-5.6 Luna is priced at $1.00 in · $6.00 out / 1M tokens — a blended rate of about $2.50 per 1M tokens for typical input-heavy agent use.
What is GPT-5.6 Luna's context window?
GPT-5.6 Luna has a 1.1M-token context window — large enough to hold sizeable codebases or document sets in a single run. OpenAI is the maker.
Can I self-host GPT-5.6 Luna?
No — GPT-5.6 Luna is a proprietary, hosted-only model from OpenAI. You access it through an API (e.g. OpenRouter) rather than running the weights yourself.
What are the downsides of GPT-5.6 Luna?
The main trade-offs: clear quality step below terra and sol; very new — behavior under long agent chains still settling. It's strongest for cheap, fast subtasks, routing, and subagents in openai-based stacks.

Luna compared

Luna vs Haiku 4.5Luna vs DeepSeek V4 FlashLuna vs GPT-5.4 MiniLuna vs Qwen3 235BLuna vs Flash LiteLuna vs Llama 4 MaverickLuna vs CodestralLuna vs Sol

The model is half the story — the agent is the other half

The model picks the moves; the agent runs the loop, the tools, and the guardrails. Once you've chosen a model, see which agent gets the most out of it.

Compare AI agents →