Gemini 3.8 Flash vs GPT-5.6 Luna
Gemini 3.8 Flash comes out ahead on 11 of 13 shared measures, and the two rank 13 and 44 of 54 overall. GPT-5.6 Luna is 3.3× cheaper per million tokens (blended) — A$1.09 vs A$3.78 per 1,000 replies of customer support replys.
Where each one wins.
Gemini 3.8 Flash ranks 13th of 54; GPT-5.6 Luna ranks 44th.
Gemini 3.8 Flash ranks 12th of 42; GPT-5.6 Luna ranks 26th.
Gemini 3.8 Flash ranks 10th of 41; GPT-5.6 Luna ranks 28th.
Gemini 3.8 Flash ranks 15th of 54; GPT-5.6 Luna ranks 32nd.
GPT-5.6 Luna ranks 18th of 54; Gemini 3.8 Flash ranks 22nd.
GPT-5.6 Luna ranks 11th of 53; Gemini 3.8 Flash ranks 17th.
GPT-5.6 Luna is 3.3× cheaper per million tokens (blended) — A$1.09 vs A$3.78 per 1,000 replies of customer support replys.
Gemini 3.8 Flash writes 1.7× faster (244 tok/s vs 145 tok/s).
Both can be used from Australia through their makers’ own APIs (Google and OpenAI), which may process requests outside Australia. Gemini 3.8 Flash can be called from Google Vertex AI in Sydney, but only through global routing, so requests may be processed outside Australia. 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.
Side by side.
“Level” means the gap is inside LMArena’s confidence interval, or the scores are identical. Independent test scores are for each model’s headline setting. Both models can be used from Australia through their makers’ APIs; the Australian rows show whether requests can also be processed in Australia.
| BENCHMARK | GEMINI 3.8 FLASH | GPT-5.6 LUNA | LEADS |
|---|---|---|---|
| AA Intelligence Index | 40.9 | 37.3 | Gemini 3.8 Flash |
| AA Coding Index | 76.3 | 71.4 | Gemini 3.8 Flash |
| GPQA Diamond | 95.3% | 91.1% | Gemini 3.8 Flash |
| Humanity’s Last Exam | 47.8% | 39.5% | Gemini 3.8 Flash |
| AA-LCR | 81.3% | 83.7% | GPT-5.6 Luna |
| Terminal-Bench 2.1 | 87.6% | 80.9% | Gemini 3.8 Flash |
| τ²-Bench (banking) | 44.9% | 31.1% | Gemini 3.8 Flash |
| SciCode | 56.6% | 53.6% | Gemini 3.8 Flash |
| LMArena Text | 1494 | 1453 | Gemini 3.8 Flash |
| LMArena WebDev | 1583 | 1518 | Gemini 3.8 Flash |
| LMArena Vision | 1286 | 1259 | Gemini 3.8 Flash |
| LMArena Document | 1462 | ||
| LMArena Agent | #19 | #31 | Gemini 3.8 Flash |
| Blended price (AUD) | A$2.16 | A$0.65 | GPT-5.6 Luna |
| Output speed | 244 tok/s | 145 tok/s | Gemini 3.8 Flash |
| Answer starts after | 14 s | 1.5 min | Gemini 3.8 Flash |
| Context window | 1,048,576 tokens | 1,050,000 tokens | GPT-5.6 Luna |
| USE FROM AUSTRALIA | |||
| Maker’s API | YES | YES | |
| Processed in Australia | NO | NO | |
| AUSTRALIAN CLOUD REGIONS | |||
| AWS Bedrock · Sydney | NOT OFFERED | GLOBAL | |
| AWS Bedrock · Melbourne | NOT OFFERED | GLOBAL | |
| Azure · Australia East | NOT OFFERED | GLOBAL | |
| Google Vertex AI · Sydney | GLOBAL | NOT OFFERED | |
The same jobs, in Australian dollars.
List prices at the snapshot’s RBA rate, excluding GST and reasoning tokens.
| JOB | GEMINI 3.8 FLASH | GPT-5.6 LUNA | MEDIAN MODEL |
|---|---|---|---|
| Customer support replyPER 1,000 REPLIES | A$3.78 | A$1.09 | ≈ A$8.20 |
| Summarise a 30-page documentPER 100 DOCUMENTS | A$2.59 | A$0.71 | ≈ A$5.35 |
| Agentic coding taskPER 10 TASKS | A$2.05 | A$0.57 | ≈ A$4.27 |
Each model in full.
Gemini 3.8 Flash or GPT-5.6 Luna?
Is Gemini 3.8 Flash better than GPT-5.6 Luna?
On our overall leaderboard Gemini 3.8 Flash ranks 13th and GPT-5.6 Luna 44th of 54. Which is better depends on the job: the table above compares each benchmark, price and Australian availability.
Which is cheaper, Gemini 3.8 Flash or GPT-5.6 Luna?
GPT-5.6 Luna, at A$0.65 per million tokens (blended) against A$2.16 for Gemini 3.8 Flash. Reasoning-heavy settings add output tokens, so real bills depend on the effort level you run.
Can I use Gemini 3.8 Flash and GPT-5.6 Luna in Australia?
Both can be used from Australia through their makers’ own APIs (Google and OpenAI), which may process requests outside Australia. Gemini 3.8 Flash can be called from Google Vertex AI in Sydney, but only through global routing, so requests may be processed outside Australia. 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.
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 →