GitHub Copilot · usage-based billing
What a session costs, per model.
Session profile
The profile is the session of the hungriest model. Every other model uses less in proportion to its Artificial Analysis tokens per coding task at high reasoning.
| Model | In / cached / out per 1M | Top AA score | Cost per session | AI credits | Sessions per 10,000 credits |
|---|---|---|---|---|---|
| GPT-6 Luna | $0.10 / $0.010 / $0.50 | 27.8 | $0.050 | 5.0 | 1,994 |
| Claude Haiku 5.5 | $0.10 / $0.010 / $0.50 | 36.2 | $0.123 | 12 | 812 |
| GPT-6.1 Terra | $1.00 / $0.050 / $5.00 | 51.0 | $0.233 | 23 | 429 |
| GPT-6.1 Sol | $2.00 / $0.10 / $10.00 | 54.5 | $0.553 | 55 | 180 |
| Claude Sonnet 5.5 | $2.00 / $0.10 / $10.00 | 54.2 | $0.897 | 90 | 111 |
| Gemini 3.8 FlashPromo | $0.75 / $0.075 / $3.75 | 30.5 | $1.14 | 114 | 87 |
| GPT-6 Astra | $10.00 / $1.00 / $50.00 | 56.9 | $2.49 | 249 | 40 |
| Claude Opus 5.5 | $4.00 / $0.20 / $20.00 | 58.4 | $2.50 | 250 | 40 |
| Grok 4.7 | $2.00 / $0.50 / $6.00 | 35.0 | $2.62 | 262 | 38 |
| Gemini 4 ArgonPromo | $2.00 / $0.10 / $10.00 | 50.4 | $3.04 | 304 | 32 |
| Claude Fable 5.1 | $10.00 / $0.25 / $50.00 | 49.7 | $7.28 | 728 | 13 |
Uses base rates at high reasoning. Several models charge more once one request passes a context threshold (for example 272K tokens for GPT-6.x, 100K for Claude Haiku 5.5). Cache writes are not modeled. Auto model selection takes 10% off on paid plans. Top AA score is the best Artificial Analysis Intelligence Index across reasoning efforts.