SIDE BY SIDE
Compare selected AI models side by side.
Compare uses, prices, speed evidence, and version details for up to three models. Language, video, decision models and historical evaluations are shown separately.
Language & coding
| What matters | Claude Opus 5.5 ↗Anthropic | Claude Sonnet 5.5 ↗Anthropic | GPT-6 Astra ↗OpenAI |
|---|---|---|---|
| Use cases | Coding · Agents & tools · Reasoning · Documents & writing · Visual understanding | Coding · Agents & tools · Documents & writing · Visual understanding · Everyday tasks | Coding · Agents & tools · Reasoning · Documents & writing · Visual understanding |
| Model score | 58Claude Opus 5.5 · max, default fallbackArtificial Analysis · Claude Opus 5.5 · max, default fallback ↗ | 56Claude Sonnet 5.5 · max, default fallbackArtificial Analysis · Claude Sonnet 5.5 · max, default fallback ↗ | 53GPT-6 Astra · maxArtificial Analysis · GPT-6 Astra · max ↗ |
| Output / 1M tokens · USD | $4 | $2 | $10 |
| Published rates · USD | $4 input / $20 outputPer 1M native tokens | $2 input / $10 outputPer 1M native tokens | $10 input / $50 outputPer 1M native tokens |
| Cached input / 1M | $0.2 | $0.2 | $1 |
| Cache write / 1M | $5 | $2.5 | $12.5 |
| Pricing conditions | Claude API · standard processingCache write shown for 5-minute TTL; 1-hour writes cost $8/MTok. Fast mode has separate pricing.Claude API pricing ↗ | Claude API · standard processingCache write shown for 5-minute TTL; 1-hour writes cost $4/MTok.Claude API pricing ↗ | OpenAI API · Standard · short contextLong-context and faster processing tiers have separate rates. Tool charges are additional; ChatGPT subscriptions are separate.OpenAI API pricing ↗ |
| Output speed · tokens/s | Not benchmarkedAnthropic's relative latency category within the Claude family; no cross-provider measurement.Claude model overview ↗ | Not benchmarkedAnthropic describes Sonnet 5.5 latency as Fast relative to its own models.Claude model overview ↗ | Not benchmarkedNo speed classification verified for this version and serving tier. This does not mean it is slow. |
| Context window | 1,000,000 | 1,000,000 | 1,050,000 |
| Model ID | claude-opus-5-5 | claude-sonnet-5-5 | gpt-6-astra |
| Version notes | Uses adaptive thinking. Historical Opus 4.6 results do not evaluate this release. | Supports adaptive thinking; the provider's speed label is not our benchmark. | The official name is Astra. Reasoning effort and processing tier affect cost and completion time. |
| Official documentation | Claude model overview ↗Checked 2026-10-05 | Claude model overview ↗Checked 2026-10-05 | GPT-6 Astra model ↗Checked 2026-10-05 |
Language prices are USD per million tokens; input and output are billed separately. Video samples use 5s of 768P output where verified. Extra tools, reasoning, cache writes, input media, storage, and taxes can change the total. (fast) marks a published median output speed above 150 tokens/s in the linked test settings.