Migrating from kimi-k2.6 to kimi-k3
What changes when you move from kimi-k2.6 to kimi-k3: input price +215.8%, output price +275%.
- From
- kimi-k2.6
- Status
- Active
- Shutdown
- none announced
- Time left
- —
- To
- kimi-k3
- Status
- Active
- Shutdown
- none announced
- Time left
- —
Prices
USD per 1M tokens, standard tier| Type | kimi-k2.6 | kimi-k3 | Change |
|---|---|---|---|
| Input | $0.95first seen October 5, 2026 · source | $3.00first seen October 5, 2026 · source | +215.8% |
| Cached input | $0.16first seen October 5, 2026 · source | $0.30first seen October 5, 2026 · source | +87.5% |
| Output | $4.00first seen October 5, 2026 · source | $15.00first seen October 5, 2026 · source | +275% |
At 1M input and 250K output tokens a month: $1.95 on kimi-k2.6, $6.75 on kimi-k3 (+246.2%).
Other tiers, prompt-size bands and modalities are on each model's page.
Limits
TokensNeither company publishes a context window or output limit for these models.
Benchmarks
Same benchmark and source on both sides; settings named where they differ| Benchmark | kimi-k2.6 | kimi-k3 | Change |
|---|---|---|---|
| Arena TextArena, max effort (vs no setting). Ranges don't overlap. | 1461 ±4 | 1488 ±5 | +27 |
| Arena Text: CodingArena, max effort (vs no setting). Ranges don't overlap. | 1516 ±7 | 1539 ±7 | +23 |
| Arena WebDevArena, max effort (vs no setting). Ranges don't overlap. | 1509 ±7 | 1655 ±6 | +146 |
| FrontierMath Tier 4Epoch AI, max effort (vs no setting). | 25.6% | 39.0% | +13.4 pts |
| FrontierMath Tiers 1-3Epoch AI, max effort (vs no setting). | 57.2% | 72.2% | +15.0 pts |
GPQA DiamondEpoch AI, max effort (vs no setting).Other settings
| 90.8% | 93.1% | +2.3 pts |
OTIS Mock AIME 2024-2025Epoch AI, max effort (vs no setting).Other settings
| 96.1% | 97.2% | +1.1 pts |
| SimpleQA VerifiedEpoch AI, max effort (vs no setting). | 34.9% | 50.6% | +15.7 pts |
A score from one source is never compared with another source's, and company-reported scores are compared only within the same company; settings named where they differ. Data: Arena (arena.ai), CC BY 4.0; Epoch AI, Capabilities & benchmarking (epoch.ai/benchmarks), CC BY.
Features and parameters
Compare modelsNeither company publishes feature facts for these models.
Aliases
All aliasesNo aliases point at either model.
On cloud platforms
All platform dates| Model | Platform | Platform ID | Shutdown | Time left |
|---|---|---|---|---|
| kimi-k2.6 | Microsoft Foundryversion 2026-04-20 | Kimi-K2.6 | Apr 16, 2027Active | 183 days |
| kimi-k3 | Amazon Bedrock | moonshotai.kimi-k3 | none announcedActive |
Read daily from each platform's own pages. The status above uses each company's own date. Time left counts to 00:00 UTC on the shutdown date, which is the evening before in ET.
Checklist
- Change the model string from
kimi-k2.6tokimi-k3. - Re-check costs: kimi-k3 costs more than kimi-k2.6 (input +215.8%, output +275%).
Compared from AIWire's published data when this page loaded. Each fact links to the page it came from. Model pages: kimi-k2.6, kimi-k3.