Side-by-side comparison · models
GPT-6 Astra vs GPT-6 Luna (batch)
What we have checked about each, line by line. Every value shows where it came from and when we last confirmed it.
Models
GPT-6 Astra
GPT-6 Astra is OpenAI's flagship model for demanding end-to-end work.
Models
GPT-6 Luna (batch)
GPT-6 Luna is the fast, cost-efficient model in OpenAI's GPT-6 series, positioned below GPT-6 Sol.
Checked facts
Line by line
| Measure | GPT-6 Astra | GPT-6 Luna (batch) |
|---|---|---|
| Input price | $10.00 per million tokens | Not checked yet |
| Output price | $50.00 per million tokens | Not checked yet |
| Price on amazon-bedrock | $1.00 per million cached input tokens | Not checked yet |
| Price on amazon-bedrock | $10.00 per million input tokens | Not checked yet |
| Price on amazon-bedrock | $50.00 per million output tokens | Not checked yet |
| Price on azure | $1.00 per million cached input tokens | Not checked yet |
| Price on azure | $10.00 per million input tokens | Not checked yet |
| Price on azure | $50.00 per million output tokens | Not checked yet |
| Price on openai | $0.500 per million cached input tokens | Not checked yet |
| Price on openai | $5.00 per million input tokens | Not checked yet |
| Price on openai | $25.00 per million output tokens | Not checked yet |
| Latest version | openai/gpt-6-astra-20260903 | openai/gpt-6-luna-20260922 |
| Time to first token (ms) | 407,700 | 122,110 |
| Time to first token (ms) on amazon-bedrock | 2,282 | Not checked yet |
| Time to first token (ms) on azure | 7,326 | Not checked yet |
| Time to first token (ms) on openai | 6,867 | Not checked yet |
| Output speed (tokens a second) | 46.1 | 120.9 |
| Output speed (tokens a second) on amazon-bedrock | 51.5 | Not checked yet |
| Output speed (tokens a second) on azure | 39 | Not checked yet |
| Output speed (tokens a second) on openai | 34 | Not checked yet |
| Tokens served on OpenRouter in a day | 32.3B (for 7 Nov 2026) | Not checked yet |
| Tool calls on OpenRouter in a day | 355.4K (for 7 Nov 2026) | Not checked yet |
| Artificial Analysis Intelligence Index | 56.9 | Not checked yet |
| Artificial Analysis Coding Index | 77.1 | Not checked yet |
| Artificial Analysis Agentic Index | 51 | Not checked yet |
| LMArena text rating (Elo) | 1,441 | Not checked yet |
| LMArena web development rating (Elo) | 1,770 | Not checked yet |
| LMArena vision rating (Elo) | 1,280 | Not checked yet |
| Gpqa diamond | 95.8% (for 3 Sept 2026) | Not checked yet |
| Frontiermath tiers 1 3 v2 private | 93.7% (for 3 Sept 2026) | Not checked yet |
| Simpleqa verified | 75.6% (for 3 Sept 2026) | Not checked yet |
| Arc agi 2 | 95% (for 3 Sept 2026) | Not checked yet |
| Apex agents | 64.7% (for 3 Sept 2026) | Not checked yet |
| Arc agi | 98.5% (for 3 Sept 2026) | Not checked yet |
| Balrog | 68.3% (for 3 Sept 2026) | Not checked yet |
| Chess puzzles | 76% (for 3 Sept 2026) | Not checked yet |
| Dtbench | 97.3% (for 3 Sept 2026) | Not checked yet |
| Ebr bench | 76.2% (for 3 Sept 2026) | Not checked yet |
| Frontiercode | 53.3% (for 3 Sept 2026) | Not checked yet |
| Frontiermath tier 4 v2 private | 97.6% (for 3 Sept 2026) | Not checked yet |
| Frontierswe | 65.5% (for 3 Sept 2026) | Not checked yet |
| Furniture assembly | 80% (for 3 Sept 2026) | Not checked yet |
| Gso bench | 79.4% (for 3 Sept 2026) | Not checked yet |
| Lmca | 64.4% (for 3 Sept 2026) | Not checked yet |
| Mirrorcode | 46.7% (for 3 Sept 2026) | Not checked yet |
| Mystery game puzzles | 84% (for 3 Sept 2026) | Not checked yet |
| Proofbench | 99% (for 3 Sept 2026) | Not checked yet |
| Remote labor index | 20.8% (for 3 Sept 2026) | Not checked yet |
| Spireclimb | 25.7% (for 3 Sept 2026) | Not checked yet |
| Weirdml | 93.6% (for 3 Sept 2026) | Not checked yet |
| Weirdml v3 | 54% (for 3 Sept 2026) | Not checked yet |
| Otis mock aime 2024 2025 | 100% (saturated benchmark: the top models all score near the ceiling) (for 3 Sept 2026) | Not checked yet |
A line appears only when at least one record has a value we confirmed from a source. Values we have not confirmed are left out, never guessed. Records of different types share fewer lines.
Add a third
Compare with one more record
How this works
Share it, or build your own
This page works for any two or three records in the directory, of any type. The link stays the same whichever order you name them in, so you can share it as it is. Change the records in the address, for example /directories/vs/claude-code-vs-cursor, or start from our editorial comparisons.