If you are wiring kimi-k3.1-pro into a coding agent or a long-context repair loop, the rate that matters right now is the promotional pair on AIHubMix: $1.80 / 1M input tokens and $9.00 / 1M output tokens, marked through 2026-11-23. The regular rates on the same model page are $2.00 / $10.00. That is about 10% off, and both sides of the comparison are listed there.
Context window: 1,048,576 tokens. Model ID: kimi-k3.1-pro. Currency: USD. This note is for budgeting, not a benchmark.
Announcement: X post.
Rate card
| Meter | Regular | Promo | Δ / 1M |
|---|---|---|---|
| Input | $2.00 | $1.80 | $0.20 |
| Output | $10.00 | $9.00 | $1.00 |
| Cache read | $0.20 | $0.18 | $0.02 |
regular = in/1e6*$2 + out/1e6*$10 promo = in/1e6*$1.8 + out/1e6*$9 | Usage | Regular | Promo | Save |
|---|---|---|---|
| 100K + 100K | $1.20 | $1.08 | $0.12 |
| 1M + 100K | $3.00 | $2.70 | $0.30 |
| 1M + 1M | $12.00 | $10.80 | $1.20 |
Estimates ignore cache hits, minimum units, and rounding.
Worked coding load
Cross-file edit: 500K in (sources, interfaces, failing tests, error logs) + 50K out (patch notes and follow-ups).
- Regular:
0.5*$2 + 0.05*$10 = $1.50 - Promo:
0.5*$1.8 + 0.05*$9 = $1.35 - Delta: $0.15
Re-run the arithmetic on your own token counters. Cache-heavy sessions need the cache-read line as well.
Minimal call
curl -s https://aihubmix.com/v1/chat/completions \ -H "Authorization: Bearer $AIHUBMIX_API_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"kimi-k3.1-pro","messages":[{"role":"user","content":"Hello"}]}' A ~1M window helps when you are evaluating repo-scale context, long-log triage, or agents that keep tool traces in-session. It is capacity, not free quota.
What to verify before you scale
- Promo window ends 2026-11-23 on the model page.
- Regular vs promo are both published rates, not a reverse-engineered list price.
- Account quotas live in the console; this post does not claim unlimited access.
