GPT-6 Astra
Intelligence
51.05
#13 of 520 models
Output speed
50.5 tok/s
Measured generation rate
Cost / index task
$1.721
Measured evaluation workload
End-to-end latency
58ms
First answer 48ms
01 / Position
A model in the landscape
Compare this model with peers across intelligence, speed, cost, and individual evaluations.
02 / Evidence
Benchmark results
21 distinct evaluations grouped by topic. Explore cross-model charts or switch to the full results list.
Domain capability
03 / Operation
Speed, tokens & price
Prices are per million tokens; latency and task costs describe the captured evaluation rather than a universal provider guarantee.
Token pricing
- Input / 1M
- $10.00
- Output / 1M
- $50.00
- Cache read / 1M
- $1.00
- Cache write / 1M
- $12.50
- Cache-hit discount
- 90%
Response & usage
- Output speed
- 50.5 tok/s
- First chunk
- 48ms
- First answer
- 48ms
- Output / task
- 11.8k
- Reasoning / task
- 4.7k
Evaluation workload
- Canonical input
- 737.87M
- Canonical output
- 26.19M
- Answer / task
- 7.1k
- Index time / task
- 4m
- Run cost
- $2,916.607
By prompt type
| Prompt | Output speed | First chunk | First answer | End to end |
|---|---|---|---|---|
| Medium | 54.2 tok/s | 27ms | 27ms | 36ms |
| Long | 50.5 tok/s | 48ms | 48ms | 58ms |
| Hundred K | 55.9 tok/s | 32ms | 32ms | 41ms |
| Medium Parallel | 46.1 tok/s | 26ms | 26ms | 37ms |
04 / Access
License & availability
Recorded license
No license recorded
No direct license document is available in the captured record. Weight availability alone does not establish usage rights.
- Weight access
- Closed / unavailable
- Classification
- Proprietary
- Reasoning
- Yes
Routing and providers
- OpenRouter ID
- openai/gpt-6-astra
- Tokenizer
- GPT
- Instruct type
- —
- Captured endpoints
- 5
| Provider | Quantization | Context | tok/s |
|---|---|---|---|
| OpenAI | unknown | 1.05M | — |
| Azure | unknown | 1.05M | — |
| OpenAI | unknown | 1.05M | — |
| Azure | unknown | 1.05M | — |
| OpenAI | unknown | 1.05M | — |
05 / Audit
Underlying fields
Inspect the captured model and routing records. Artificial Analysis destination links are omitted.