GPT-5.4 Mini vs GPT-5.6 Luna
Benchmark Performance
Available Benchmarks
| Benchmark | GPT-5.4 Mini | GPT-5.6 Luna |
|---|---|---|
| ARC-AGI-1verified-v1-b2e2a31b3d53 · verified_score · leader | 63.6770% of row best · percent · GPT-5.4 Mini (XHigh) | 90.67100% of row best · percent · GPT-5.6 Luna 2026-07-30 (Max) |
| ARC-AGI-2verified-v2-326661568d5f · verified_score · leader | 18.9032% of row best · percent · GPT-5.4 Mini (XHigh) | 59.58100% of row best · percent · GPT-5.6 Luna 2026-07-30 (Max) |
| LMArena Text Arenatext-2026-09-01-011508720696 · arena_rating · leader | 1,412.1099% of row best · rating · gpt-5.4-mini-high; 95% CI [1408.28461201, 1415.92470838]; votes 59472; rank 123 | 1,430.14100% of row best · rating · gpt-5.6-luna-xhigh; 95% CI [1425.14784950, 1435.13857729]; votes 24433; rank 82 |
| LMArena Vision Arenavision-2026-08-27-011508720696 · arena_rating · statistical tie | 1,245.4899% of row best · rating · gpt-5.4-mini-high; 95% CI [1239.01417013, 1251.95452638]; votes 24076; rank 54 | 1,254.30100% of row best · rating · gpt-5.6-luna-xhigh; 95% CI [1245.05722338, 1263.53552757]; votes 5857; rank 45 |
| LiveBench2026-06-25 · overall · leader | 69.5390% of row best · percent · gpt-5.4-mini-xhigh · 47,448 output tokens / case | 77.05100% of row best · percent · gpt-5.6-luna-max · 21,799 output tokens / case |
| ToneBench2026-08-28-10-task-cd9819ab6e4d · overall_score · leader | 80.4499% of row best · points · GPT-5.4 mini · 2,559 output tokens / case | 81.47100% of row best · points · GPT-5.6 Luna (default) · 3,295 output tokens / case |
| Overall ResultCounted from the protocol-matched rows above · 1 tie | 0 benchmark wins | 5 benchmark winsOverall lead |
Third-party benchmark Only like-for-like primary-publisher results are shown; raw scores, relative scores, configuration, and token spend remain visible.
Technical Differences
Side-by-Side Facts
| Field | GPT-5.4 Mini | GPT-5.6 Luna |
|---|---|---|
| Developer | OpenAI | OpenAI |
| Family | Gpt 5 4 | Gpt 5 6 |
| Model | GPT-5.4 Mini | GPT-5.6 Luna |
| Version | GPT-5.4 Mini | GPT-5.6 Luna |
| Lifecycle | active | active |
| Released | 2026-03-17 | Unknown |
| Knowledge cutoff | 2025-08-31 | 2026-02-16 |
| Input modalities | Text, Image | Text, Image |
| Output modalities | Text | Text |
| Context window | 400,000 | 1,050,000 |
| Total parameters | Unknown | Unknown |
| Active parameters | Unknown | Unknown |
| License | Unknown | Unknown |
| Open weights | No | No |
| API available | Yes | Yes |
| Self-hostable | No | No |
| Provider access | Openai (Standard), Openrouter (Standard) | Openai (Standard), Openrouter (Standard) |
| Capabilities | chat, generation, reasoning, structured_outputs, tools | chat, generation, reasoning, tools |
14 comparable fields · 6 material differences · Pair passes the primary-source comparison gate
GPT-5.4 Mini Capabilities
GPT-5.6 Luna Capabilities
Internal Comparison Graph
Related Comparisons
| A | Pair | B | Context |
|---|---|---|---|
GPT-5 MiniOpenAI | vs | GPT-5.4 MiniOpenAI | family variantsimage, text |
GPT-5.2 ProOpenAI | vs | GPT-5.4 MiniOpenAI | family variantsimage, text |
GPT-5.4 MiniOpenAI | vs | GPT-5.5 ProOpenAI | family variantsimage, text |
 Gemma 4 26B-A4BGoogle DeepMind | vs | GPT-5.4 MiniOpenAI | cross-developer peersimage, text |
Claude Opus 4.5Anthropic | vs | GPT-5.4 MiniOpenAI | cross-developer peersimage, text |
GPT-5.4 MiniOpenAI | vs | cross-developer peersimage, text | |
GPT-5.4 ProOpenAI | vs | GPT-5.6 LunaOpenAI | family variantsimage, text |
GPT-4.1OpenAI | vs | GPT-5.6 LunaOpenAI | family variantsimage, text |
 MiniMax-M3MiniMax | vs | GPT-5.6 LunaOpenAI | cross-developer peersimage, text |
Claude Fable 5Anthropic | vs | GPT-5.6 LunaOpenAI | cross-developer peersimage, text |
Claude Opus 4.8Anthropic | vs | GPT-5.6 LunaOpenAI | cross-developer peersimage, text |
Claude Sonnet 5Anthropic | vs | GPT-5.6 LunaOpenAI | cross-developer peersimage, text |
Primary Evidence
Sources and Freshness
Questions
GPT-5.4 Mini vs GPT-5.6 Luna FAQs
Is GPT-5.4 Mini or GPT-5.6 Luna better for coding?+
This comparison does not currently contain a protocol-matched coding benchmark for both GPT-5.4 Mini and GPT-5.6 Luna, so Model Markets cannot name a coding leader from pricing, context size, or capability labels alone.
Which is cheaper, GPT-5.4 Mini or GPT-5.6 Luna?+
GPT-5.4 Mini is $0.75 and GPT-5.6 Luna is $0.20 per million tokens, so GPT-5.6 Luna is cheaper on this metric. GPT-5.4 Mini is $4.50 and GPT-5.6 Luna is $1.20 per million tokens, so GPT-5.6 Luna is cheaper on this metric.
Which has a larger context window, GPT-5.4 Mini or GPT-5.6 Luna?+
GPT-5.6 Luna has the larger sourced context window. GPT-5.4 Mini supports 400,000 and GPT-5.6 Luna supports 1,050,000.
Which performs better in benchmarks, GPT-5.4 Mini or GPT-5.6 Luna?+
GPT-5.6 Luna leads the current overall benchmark count. The result uses 6 protocol-matched benchmarks from 4 publishers; it is not a universal quality score.
Can GPT-5.4 Mini or GPT-5.6 Luna be self-hosted?+
Both models have the same recorded self-hosting status: unsupported. GPT-5.4 Mini is not marked open weight; GPT-5.6 Luna is not marked open weight.
Can GPT-5.4 Mini and GPT-5.6 Luna understand images?+
GPT-5.4 Mini is documented with image input; GPT-5.6 Luna is documented with image input. This reflects supported input modalities, not vision quality.
Which can generate longer answers, GPT-5.4 Mini or GPT-5.6 Luna?+
Neither has a larger sourced maximum output. GPT-5.4 Mini is 128,000 and GPT-5.6 Luna is 128,000.
Do GPT-5.4 Mini and GPT-5.6 Luna support reasoning and tool use?+
GPT-5.4 Mini: reasoning, tool calling, and image input. GPT-5.6 Luna: reasoning, tool calling, and image input. Feature support does not establish relative quality.
Which is available from more inference providers, GPT-5.4 Mini or GPT-5.6 Luna?+
GPT-5.4 Mini has 2 sourced provider routes; GPT-5.6 Luna has 2, a tie.
Which offers better value, GPT-5.4 Mini or GPT-5.6 Luna?+
There is no universal value winner. Compare the input and output prices above with the matched benchmark result for your workload: cheaper tokens can be offset by different quality, token usage, latency, or provider availability.