Kimi K2.6 vs Kimi K3
Kimi K2.6 (Kimi) versus Kimi K3 (Kimi): API pricing, context window, capabilities and uptime, from the same data as each model page, at the WandAPI default channel price. Input costs 3.9 Credits / M for Kimi K2.6 and 12 Credits / M for Kimi K3, so Kimi K2.6 is 68% cheaper. Context windows are 256K and 1M.
Specs side by side
| Item |  Kimi K2.6kimi-k2.6 | Kimi K3kimi-k3 |
|---|---|---|
| Input (prompt) · Credits / M | 3.9≈ $0.593 | 12≈ $1.82 |
| Output (completion) · Credits / M | 16.2≈ $2.46 | 60≈ $9.12 |
| Cache read · Credits / M | 0.66 | 1.2 |
| Context window | 256K | 1M |
| Max output | — | 128K |
| 7-day uptime | — | — |
| Latest probe latency | — | — |
| Measured throughput | 2.4 t/s | 11.5 t/s |
| Protocols | Anthropic Messages | Anthropic Messages / OpenAI Responses |
| Sample monthly cost (10M in + 2M out) | 71.4≈ $10.85 | 240≈ $36.48 |
Prices are the WandAPI default channel price in Credits (1 Credit = ¥1); green marks the better side. Uptime and latency come from WandAPI probes; a dash means the model is not probed yet.
Capabilities
FAQ
Which is cheaper, Kimi K2.6 or Kimi K3?
At the WandAPI default channel price, Kimi K2.6 input costs 3.9 Credits / M and Kimi K3 input costs 12 Credits / M, so Kimi K2.6 is 68% cheaper. Output and cache prices are in the specs table above.
Which has the larger context window, Kimi K2.6 or Kimi K3?
Kimi K2.6 offers 256K tokens, Kimi K3 offers 1M tokens.
Can I switch between Kimi K2.6 and Kimi K3 at any time?
Yes. WandAPI calls every model with the same API key; switching is just changing the model field in the request, and billing is split per model in one console.
How do the capabilities of Kimi K2.6 and Kimi K3 differ?
Only Kimi K3: File Input, Web Search, Session. Every other capability is either supported by both or by neither; see the capability table above.
More comparisons
One API key for both Kimi K2.6 and Kimi K3
Free credits on sign-up, pay as you go, no monthly fee.