GPT-5.6 Luna against 49 models

GPT-5.6 Luna costs $0.88 per 1,000 chat replies at list price. Here is every model teams spend the most on or move traffic to, sorted from cheapest, with a link to each side-by-side page.

GPT-5.6 Luna compared with 49 models
ModelPer 1,000 chat repliesTo scaleAgainst GPT-5.6 LunaConfidence signal
gpt-oss-20b$0.07292% cheaperyes
Qwen3.7 Flash$0.11287% cheaperyes
Nova Micro 1.0$0.12686% cheaperno
Command R7B (12-2024)$0.13585% cheaperno
DeepSeek V4 Flash 0731$0.1781% cheaperyes
Phi 4$0.19678% cheaperno
Nova Lite 1.0$0.21675% cheaperno
Nemotron 3.5 Lightning$0.2473% cheaperyes
DeepSeek V4.1 Flash$0.2967% cheaperyes
Gemma 4 31B$0.31664% cheaperyes
Llama 3.3 70B Instruct$0.32863% cheaperyes
Ministral 3 8B 2512$0.3659% cheaperno
MiMo-V2.6-Flash$0.39255% cheaperno
GPT-6 Luna$0.4055% cheaperno
Qwen3.8 Flash$0.48845% cheaperyes
GLM 5.3 Flash$0.5043% cheaperyes
gpt-oss-120b$0.5439% cheaperyes
Mistral Small 4$0.5439% cheaperno
Llama 4 Maverick$0.63628% cheaperyes
GPT-5.4 Nano$0.90sameno
MiniMax M3$1.081.2× the priceyes
GPT-4.1 Mini$1.441.6× the priceno
Mistral Large 3 2512$1.601.8× the priceno
Gemini 3.5 Flash Lite$1.601.8× the priceno
Qwen3.8 27B$2.042.3× the priceyes
Nemotron 3 Ultra$2.162.5× the priceno
Kimi K2.6$2.663.0× the priceyes
Hy4 preview$2.673.0× the priceno
DeepSeek V4 Pro 0423$2.673.0× the priceyes
Gemini 3.8 Flash$3.003.4× the priceno
GPT-5.4 Mini$3.303.8× the priceno
Claude Haiku 4.5$4.004.5× the priceno
GLM 5.3$4.565.2× the priceyes
Grok 4.7$5.125.8× the priceyes
Qwen3.8 Max (0902)$6.407.3× the priceyes
Claude Sonnet 5$8.009.1× the priceno
GPT-6 Sol$8.009.1× the priceno
GPT-5.6 Sol$8.009.1× the priceno
GPT-5.6 Terra$8.8010× the priceno
Gemini 3.1 Pro Preview$8.8010× the priceno
Command A$9.0010× the priceno
Nova Premier 1.0$10.0011× the priceno
GPT-5.4$11.0013× the priceno
Kimi K3$12.0014× the priceyes
Claude Opus 5.5$16.0018× the priceno
Claude Opus 5$20.0023× the priceno
GPT-5.5$22.0025× the priceno
Claude Fable 5.1$40.0045× the priceno
GPT-6 Astra$40.0045× the priceno

Find the cheapest model that answers your GPT-5.6 Luna traffic as well

A pilot tests the candidates on your own requests.

Request a pilot