Kimi K3
Kimi K3 is served by 25 companies at prices ranging from $0.72 to $4.50 per million input tokens, a 6.3× difference for identical weights.
- cheapest host · Wafer
- $0.72
- price spread across hosts
- 6.3×
- companies serving it
- 25
- intelligence index
- 43.6
Is it worth it?
Beaten on both axes. GLM 5.3 Flash scores 41.8 against 43.6 and costs 27.2× less for the same work.
Every host, compared
The same weights, priced and configured differently by 25 companies. Precision matters: 4-bit builds measurably degrade output and cost the same.
| Host | $/1M in | $/1M out | Cached | Context | Precision | Uptime |
|---|---|---|---|---|---|---|
| Wafercheapest | $0.720 | $24.00 | $0.300 | 1.05M | unknown | 100% |
| InferenceNet | $0.730 | $10.00 | $0.550 | 1.05M | fp4 | 100% |
| Sail Research | $0.840 | $13.50 | $0.300 | 1.05M | fp4 | 51.5% |
| Morph | $1.00 | $14.90 | $0.290 | 1.05M | fp8 | 64.3% |
| AkashML | $1.20 | $14.00 | $1.20 | 1.05M | fp4 | 48.3% |
| Makora | $1.53 | $12.75 | $0.204 | 1.05M | fp4 | 100% |
| Relace | $2.00 | $14.00 | $0.300 | 1.05M | fp4 | 100% |
| Decart | $2.55 | $12.75 | $0.255 | 1.05M | mxfp4 | 24.4% |
| Phala | $2.55 | $12.75 | $0.255 | 1.05M | unknown | 90.9% |
| DigitalOcean | $2.55 | $12.95 | $0.255 | 1.05M | unknown | 100% |
| Parasail | $2.60 | $13.00 | $0.260 | 1.05M | fp4 | 100% |
| Together | $2.70 | $13.50 | $0.270 | 1.05M | unknown | 98.1% |
| Wafer | $2.80 | $15.50 | $0.300 | 1.05M | unknown | 100% |
| DeepInfra | $2.85 | $14.25 | $0.285 | 1.05M | mxfp4 | 68.4% |
| InferenceNet | $2.99 | $15.00 | $0.450 | 250K | fp4 | 100% |
| Parasail | $3.00 | $15.00 | $0.300 | 1.00M | fp4 | 99.7% |
| Amazon Bedrock | $3.00 | $15.00 | $0.300 | 1.05M | unknown | 100% |
| Chutes | $3.00 | $15.00 | $0.300 | 1.05M | mxfp4 | 77.3% |
| Modal | $3.00 | $15.00 | $0.300 | 1.05M | mxfp4 | 100% |
| Fireworks | $3.00 | $15.00 | $0.300 | 1.05M | unknown | 99.9% |
| BaseTen | $3.00 | $15.00 | $0.300 | 1.05M | fp8 | 100% |
| Moonshot AI | $3.00 | $15.00 | $0.300 | 1.05M | mxfp4 | 100% |
| Alibaba | $3.45 | $17.25 | $0.345 | 1.05M | unknown | 99.8% |
| Fireworks | $4.50 | $22.50 | $0.450 | 1.05M | unknown | 100% |
| Fireworks | $4.50 | $22.50 | $0.450 | 1.05M | unknown | 100% |
8 of 25 hosts serve this model at 4-bit precision (InferenceNet, Sail Research, AkashML, Makora, Relace, Parasail, InferenceNet, Parasail), at the same price as full-precision hosts. See why this matters.
What it costs to run
- Marketing emails
- $6.37
- per 1,000 · short brief in, one email out
- Support replies
- $6.24
- per 1,000 · ticket plus help-centre context
- Lead enrichment
- $7.85
- per 1,000 · company page in, fields out
- Long-form drafts
- $29.24
- per 1,000 · outline in, full draft out
- Questions on docs
- $26.00
- per 1,000 · one large document, many questions
- Agent / tool loop
- $25.31
- per 1,000 · long context reused each step
Priced on the median host. Change the workload on the leaderboard to see how the ranking moves.
Specification
- Context, most hosts
- 1.05M · 1,573 pages
- Context range across hosts
- 250K to 1.05M
- Max output
- 944K
- Cache discount
- 80% off cached input
- Input types
- text, image, video
- Tool use
- supported
- Reasoning
- optional
- Arena Elo
- 1354 across 7 categories
- Agentic Elo
- 1266
- AA Intelligence
- 43.6
Other models with the same host spread
Host pricing and precision read live from OpenRouter’s endpoint API, compiled 12 October 2026. Free to quote with attribution. See all findings.