Choose two or three models. We’ll keep capability, price, context, caveats, and evidence in the same view.
OpenAI’s current flagship for complex end-to-end work. Text and image inputs with 128K max output.
Standard API pricing per MTok below 272K input tokens. Requests above that threshold are $20 input / $75 output per MTok.
Public API model for natural full-duplex voice conversations. Audio and text input with audio and text output.
Voice sessions cost $0.05 per minute, billed per second. Backend model and tool usage are billed separately.
| Decision factor | GPT-6 Astra | GPT-Live-1 |
|---|---|---|
| Editorial fit40% coding · 35% reasoning · 25% tool use | 10.0Best in this set | 1.0 |
| CodingNormalized editorial dimension | 10.0Best in this set | 1.0 |
| ReasoningNormalized editorial dimension | 10.0Best in this set | 1.0 |
| Tool useNormalized editorial dimension | 10.0Best in this set | 1.0 |
| Context windowProvider specification | 1.05M | Not stated |
| Input priceUSD per 1M tokens | $10 | $null |
| Output priceUSD per 1M tokens | $50 | $null |
| Best fitFrom the catalog record |
|
|
| Source trailOpen before choosing |
Quality floor: eliminate any model that cannot reliably complete the task, regardless of price.
Operating constraint: compare real context, throughput, privacy, and deployment requirements.
Total cost: calculate the full prompt/output mix and account for retries, caching, and routing.