GPT-5.6 Luna benchmarks
GPT-5.6 Luna is a proprietary model from OpenAI, released 9 Jul 2026. It ranks 44th of 54 on our overall leaderboard and 11th of 53 for value. At A$0.65 per million tokens (blended), it costs about 6.7× cheaper than the median model we track. It writes about 145 tokens a second, and answers start after 1.5 min on average, thinking included. It can be called from AWS Bedrock in Sydney and Melbourne and Microsoft Azure in Australia East, but only through global routing, so requests may be processed outside Australia.
- CONTEXT WINDOW
- 1,050,000 tokens
- MAX OUTPUT
- 128,000 tokens
- INPUT
- text, image
- WEIGHTS
- Proprietary
- RELEASED
- 9 Jul 2026
- OUTPUT SPEED
- 145 tok/s
- ANSWER STARTS AFTER
- 1.5 min (thinking included)
GPT-5.6 Luna across our rankings.
Benchmark results.
Independent test scores are for the max effort setting. “Among tracked” ranks GPT-5.6 Luna against the 54 models on this site; LMArena ranks run across its full leaderboard.
| BENCHMARK | RESULT | AMONG TRACKED | SOURCE DETAIL |
|---|---|---|---|
| INDEPENDENT TESTS · ARTIFICIAL ANALYSIS | |||
| AA Intelligence Index | 37.3 | #33 of 54 | max effort |
| AA Coding Index | 71.4 | #22 of 42 | max effort |
| GPQA Diamond | 91.1% | #27 of 44 | max effort |
| Humanity’s Last Exam | 39.5% | #40 of 54 | max effort |
| AA-LCR | 83.7% | #9 of 54 | max effort |
| Terminal-Bench 2.1 | 80.9% | #20 of 42 | max effort |
| τ²-Bench (banking) | 31.1% | #30 of 41 | max effort |
| SciCode | 53.6% | #30 of 49 | max effort |
| BLIND HUMAN VOTES · LMARENA | |||
| LMArena Text37,593 votes · ±4 | 1453 | #41 of 49 | #72 of 410 on LMArena · 30 Sep 2026 |
| LMArena WebDev12,545 votes · ±6 | 1518 | #38 of 52 | #46 of 137 on LMArena · 30 Sep 2026 |
| LMArena Vision10,099 votes · ±8 | 1259 | #31 of 36 | #42 of 156 on LMArena · 28 Sep 2026 |
| LMArena Document5,766 votes · ±8 | 1462 | #18 of 26 | #23 of 44 on LMArena · 13 Sep 2026 |
| LMArena Agent39,830 sessions | #31 | #29 of 42 | #31 of 46 on LMArena · 30 Sep 2026 |
What GPT-5.6 Luna costs in Australian dollars.
List prices converted at the snapshot’s RBA rate, excluding GST. Reasoning models are billed for their thinking as output tokens, so heavier settings cost more than the job estimates show.
| PER MILLION TOKENS | AUD | USD LIST |
|---|---|---|
| Input tokens | A$0.29 | US$0.20 |
| Output tokens | A$1.73 | US$1.20 |
| Blended (3 in : 1 out) | A$0.65 | US$0.45 |
| JOB | COST (AUD) | MEDIAN MODEL |
|---|---|---|
| Customer support replyPER 1,000 REPLIES | A$1.09 | ≈ A$8.20 |
| Summarise a 30-page documentPER 100 DOCUMENTS | A$0.71 | ≈ A$5.35 |
| Agentic coding taskPER 10 TASKS | A$0.57 | ≈ A$4.27 |
| SETTING | INTELLIGENCE | A$ / 1M | SPEED |
|---|---|---|---|
| max effort | 37.3 | A$0.65 | |
| xhigh effort | 34.6 | A$0.65 | |
| high effort | 32.1 | A$0.65 | |
| medium effort | 25.0 | A$0.65 | |
| low effort | 21.0 | A$0.65 | |
| Non-reasoning | 15.5 | A$0.65 |
Running GPT-5.6 Luna in Australia.
It can be called from AWS Bedrock in Sydney and Melbourne and Microsoft Azure in Australia East, but only through global routing, so requests may be processed outside Australia. How the platforms compare →
| PLATFORM | AVAILABILITY | DETAIL |
|---|---|---|
| AWS Bedrock · Sydney | GLOBAL | |
| AWS Bedrock · Melbourne | GLOBAL | |
| Azure · Australia East | GLOBAL | |
| Google Vertex AI · Sydney | NOT OFFERED |
Compare it with.
GPT-5.6 Luna, answered.
How much does GPT-5.6 Luna cost in Australian dollars?
GPT-5.6 Luna lists at US$0.20 per million input tokens and US$1.20 per million output tokens — A$0.29 and A$1.73 at A$1 = US$0.6949 (RBA, 1 Oct 2026), excluding GST. A typical customer support reply works out at about A$1.09 per 1,000 replies, before any reasoning tokens.
Can I use GPT-5.6 Luna in Australia with data kept onshore?
GPT-5.6 Luna can be called from AWS Bedrock in Sydney and Melbourne and Microsoft Azure in Australia East, but only through global routing, so requests may be processed outside Australia. Availability comes from the cloud providers’ own documentation; check your provider’s current terms before relying on it for data-residency obligations.
Is GPT-5.6 Luna good for coding?
It ranks 26th of 42 on our coding ranking, with an Artificial Analysis Coding Index of 71.4 and an LMArena WebDev rating of 1518 (46th of 137). The current leader is Claude Fable 5.1.
How good is GPT-5.6 Luna at agent and automation work?
It ranks 28th of 41 on our agents ranking, completing 31.1% of τ²-Bench banking customer-service tasks and 80.9% of Terminal-Bench 2.1 tasks.
How fast is GPT-5.6 Luna?
Artificial Analysis measures GPT-5.6 Luna at about 145 tok/s of output, with the answer starting after 1.5 min on average once thinking time is included (max effort setting).
What is GPT-5.6 Luna’s context window?
OpenAI documents a context window of 1,050,000 tokens (1.05M), with up to 128,000 output tokens. On AA-LCR, which tests reasoning across ~100,000-token document sets, it scores 83.7%.
Ratings: LMArena leaderboard dataset (CC BY 4.0), rescaled for the lens scores · leaderboards to 30 Sep 2026
Evaluations, prices and speed: Artificial Analysis (artificialanalysis.ai) · fetched 1 Oct 2026
Exchange rate: Reserve Bank of Australia, table F11.1 · A$1 = US$0.6949 on 1 Oct 2026 · prices exclude GST
Snapshot 2 Oct 2026 · updated weekly · How the rankings work →