Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Compare
Which of these fits your task and budget? See where each is stronger, what it costs and how fast it answers. Add up to four models to widen the comparison.
Add the two to four models you are deciding between, then press Compare.
Selected models 2 of 4
Start with where each model is stronger, check its fact sheet, then weigh value, speed and cost. Every figure links to its source; one no source published reads Unavailable.
The further a shape reaches along an axis, the better that model scored there. Pick the one that leads on the capabilities your task needs.
One card per model: price, context, speed and every capability score at a glance. Open Full evidence for the sources behind each figure.
Evidence key: ObservedUnavailable
Score against the 75% input / 25% output blended token price. Models on the line give the highest score for what they cost; a model below it costs more for the same score or less.
Time to first token is the wait before an answer starts; throughput is how fast it streams after that. A shorter wait and a higher throughput feel faster.
Time to first token: latency to the first answer chunk (Artificial Analysis, via BenchLM). Reasoning models include thinking time, so values can run to tens or hundreds of seconds.
Input and output prices per million tokens, plus a 75% input / 25% output blend for a like-for-like figure. For a monthly estimate, use the cost simulator.
The numbers behind the charts, for a final check before you choose: cache and batch prices, cost per successful task, context window and release date.
| Metric | GPT-6.2 Astra | Gemini 4 Carbon |
|---|---|---|
| Input / 1M | $8.00 | $2.00 |
| Output / 1M | $40.00 | $10.00 |
| Cache read / 1M | $0.40 | $0.10 |
| Cache write / 1M | $10.00 | $0.19 |
| Blended / 1M (75% input / 25% output) | $16.00 | $4.00 |
| Cost per successful task (LiveBench) | Unavailable | Unavailable |
| Batch input / 1M (OpenRouter batch) | $4.00 | $1.00 |
| Batch output / 1M (OpenRouter batch) | $20.00 | $5.00 |
| Batch cache read / 1M (OpenRouter batch) | $0.20 | $0.05 |
| Batch cache write / 1M (OpenRouter batch) | $5.00 | $0.19 |
| Long-context batch price tiers (OpenRouter batch) | Above 272K prompt tokens: $8.00 input / $30.00 output per 1M · cache read $0.40 · cache write $10.00 | Unavailable |
| Time to first token | Unavailable | 71.28 s |
| Throughput | 39 tok/s | 55 tok/s |
| Context window | 1.05M tokens | 1M tokens |
| Release date | 2026-11-17 | 2026-10-28 |
| Access | Proprietary | Proprietary |
Batch rows: Batch API rate (asynchronous jobs), from OpenRouter's batch listing.
For each pair the comparison source has tested: which model wins each capability, and by how much. A gap the source did not publish stays Unavailable.
The comparison source only covers pairs it has evaluated. The fact sheets and charts above still compare every model on the same published figures.