MoonshotAI: Kimi K2.6
moonshotai/kimi-k2.6
Cheapest provider
$0.68 / 1M
Baidu
Fastest provider (p95)
—
No throughput data yet — populated as traffic accumulates
Intelligence (composite)
87.8
MMLU-Pro · HumanEval · math · GPQA
Per-provider performance
Latency / throughput / uptime / price measured across providers over the last 30 minutes of live traffic. This is what proves “sourced cheapest” — Atlas mode draws on these per call to serve the cheapest path that holds quality.
| Provider | Quant | Input $/1M | Output $/1M | Latency p50 / p95 | Throughput p50 / p95 | Uptime 30m | Success |
|---|---|---|---|---|---|---|---|
| Baidu | q4· fp4 | $0.6840 | $3.4200 | — | — | 98.73% | — |
| Inceptron | q4· int4 | $0.7300 | $3.5000 | — | — | 98.91% | — |
| Io Net | q4· int4 | $0.7300 | $3.4900 | — | — | 98.58% | — |
| Chutes | q4· int4 | $0.7400 | $3.5000 | — | — | 98.52% | — |
| Cloudflare | undisclosed | $0.7400 | $3.5000 | — | — | 99.46% | — |
| DeepInfra | q4· fp4 | $0.7500 | $3.5000 | — | — | 99.68% | — |
| Parasail | q4· int4 | $0.7500 | $3.5000 | — | — | 99.66% | — |
| SiliconFlow | q8· fp8 | $0.7700 | $4.0000 | — | — | 100.00% | — |
| Novita | undisclosed | $0.8000 | $3.4000 | — | — | 97.22% | — |
| DigitalOcean | undisclosed | $0.8075 | $3.4000 | — | — | 99.80% | — |
| Venice | q4· int4 | $0.8500 | $4.6550 | — | — | 99.65% | — |
| StreamLake | undisclosed | $0.8550 | $3.6000 | — | — | 100.00% | — |
| AkashML | q4· int4 | $0.9500 | $4.0000 | — | — | 99.37% | — |
| AtlasCloud | q4· int4 | $0.9500 | $4.0000 | — | — | 98.40% | — |
| Fireworks | undisclosed | $0.9500 | $4.0000 | — | — | — | — |
| Moonshot AI | q4· int4 | $0.9500 | $4.0000 | — | — | 99.93% | — |
| WandB | q4· fp4 | $0.9500 | $4.0000 | — | — | 99.95% | — |
| Phala | undisclosed | $1.0900 | $4.6000 | — | — | 97.71% | — |
| Together | undisclosed | $1.2000 | $4.5000 | — | — | 99.70% | — |
“—” means live telemetry hasn’t accumulated enough recent traffic for that endpoint. “undisclosed” means the provider serves the model but doesn’t expose the quantization label (typically running fp8 / int8 internally).
Intelligence breakdown
Composite score is a weighted average of public benchmarks (30% MMLU-Pro, 25% code pass@1, 25% math, 20% GPQA). Numbers come from model cards and the Artificial Analysis intelligence harness; missing components are renormalised over what’s present.
MMLU-Pro
84.6
broad reasoning
Code
89.6
pass@1 (HumanEval / LiveCodeBench)
MATH
—
math accuracy
GPQA Diamond
90.5
hard reasoning
Source: Kimi K2.6 (MMLU-Pro 84.6, GPQA-Diamond 90.5, LiveCodeBench v6 89.6); codersera/llmreference cross-model tables + Artificial Analysis Intelligence Index 54
How Atlas mode sources MoonshotAI: Kimi K2.6
- Strict mode — pin MoonshotAI: Kimi K2.6 exactly and we pass it straight through, sourced from the cheapest provider above. The same model, no substitutions — currently Baidu at $0.68/1M.
- Atlas mode — the default. Each call is auto-optimized for the cheapest path that holds quality, at least 5% off going direct from call one and climbing as it ramps. You always see which model served the call and exactly what you saved — thumbs-down anything you don’t like for a full refund.