by Moonshot AI
Kimi K2.5 is the latest model from Moonshot AI with native multimodal design for direct image/screenshot analysis. Full-parameter RL tuning, agent swarm capabilities, and dual Thinking/Instant modes.
Parameters
1T (MoE)
Architecture
Mixture of Experts (384 experts)
Context
256K
Provider
Moonshot AI
Drop-in replacement for OpenAI API. Just change the base URL.
Only pay for actual GPU compute time. No idle costs.
99.9% uptime SLA, SOC 2 compliant, dedicated support.
Scales from zero to thousands of requests automatically.
| Fleek | Fireworks | Together | Baseten | |
|---|---|---|---|---|
| Input | $0.12 | $0.60 | $1.00 | $0.60 |
| Output | $0.47 | $2.50 | $3.00 | $2.50 |
| Savings | 70% | 70% | 70% |
Prices are per million tokens. Fleek pricing based on $0.0025/GPU-second.
See how much you'd save running Kimi K2.5 on Fleek
| Model Name | Kimi K2.5 |
| Total Parameters | 1T (MoE) |
| Active Parameters | 32B |
| Architecture | Mixture of Experts (384 experts) |
| Context Length | 256K tokens |
| Inference Speed | 34,000 tokens/sec |
| Provider | Moonshot AI |
| Release Date | Jan 27, 2026 |
| License | Modified MIT |
| HuggingFace | https://huggingface.co/moonshotai/Kimi-K2.5-Instruct |
Software Engineering benchmark - real GitHub issues
Click any benchmark to view the official leaderboard. Rankings among open-source models.
Join the waitlist for early access. Start free with $5 in credits.