Pricing
The price, and where to check it.
Per token, no reservation and no commitment. A rate means nothing until you know what’s behind it, which is why the precision label sits beside it here and on OpenRouter’s provider page, independently of us.
| Model | Input / 1M | Output / 1M | Cache read / 1M | Context | Action |
|---|---|---|---|---|---|
| Kimi K3.1 Pro | $1.75 | $10.00 | $0.18 | 1M | Test |
| GLM 5.3 | $0.19 | $4.40 | $0.19 | 1M | Test |
| Qwen 3.8 27B | $0.225 | $1.98 | $0.05 | 262K | Test |
| Qwen 3.6 35B A3B | $0.10 | $0.90 | $0.05 | 262K | Test |
| GPT OSS 120B | $0.037 | $0.187 | $0.037 | 128K | Test |
| GPT OSS 20B | $0.018 | $0.09 | $0.009 | 128K | Test |
| Llama 3.3 70B | $0.20 | $0.52 | $0.10 | 128K | Test |
Spend estimator
See what your agent stack actually costs.
Agentic coding loops re-read the same context on every turn, which is where the bill actually comes from. Estimate it against published list rates before you commit to anything.
Monthly token assumptions
- Input · 120M
- $27
- Output · 24M
- $48
- Cache reads · 1.4B
- $68
AkashML cost
$143/mo
First-party API
$615/mo
Claude Sonnet 5.5 list: $2 / $10 / $0.10 cache
Pricing, with the label
Every price cell carries a precision label.
The same weights, on the hosts that also run them. Where a provider does not say what precision it serves, its number and ours are not comparable, even where theirs is lower.
| Model · $/1M tokens | AkashML | DeepInfra | Together | Fireworks |
|---|---|---|---|---|
| Kimi K3.1 Pro | $1.75 in $10.00 out $0.18 cache read fp4 | $1.90 in$9.50 outbf16 | $2.00 in$10.00 outunknown | $2.00 in$10.00 outunknown |
| GLM 5.3 | $0.19 in $4.40 out $0.19 cache read fp4 | $0.45 in$2.00 outfp4 | $1.40 in$4.40 outunknown | $0.98 in$3.08 outunknown |
| GPT OSS 120B | $0.037 in $0.187 out $0.037 cache read fp4 | $0.037 in$0.17 outbf16 | $0.15 in$0.60 outunknown | not served |
| Qwen 3.8 27B | $0.225 in $1.98 out $0.05 cache read fp8 | $0.15 in$1.875 outbf16 | not served | not served |
| Qwen 3.6 35B A3B | $0.10 in $0.90 out $0.05 cache read fp8 | $0.08 in$0.76 outfp8 | not served | not served |
| GPT OSS 20B | $0.018 in $0.09 out $0.009 cache read fp4 | $0.024 in$0.112 outbf16 | not served | not served |
| Llama 3.3 70B | $0.20 in $0.52 out $0.10 cache read fp8 | $0.10 in$0.32 outfp8 | not served | not served |
Rates read Nov 18, 2026 · competitor cells link to OpenRouter
Side by side
Against the closed APIs.
First-party list rates from Anthropic and OpenAI, each linking to the vendor's own pricing page. Add or remove any column.
Open models on AkashML
Closed first-party APIs
3 of 5 columns
| Attribute | Kimi K3.1 ProAkashML | Claude Sonnet 5.5Anthropic | GPT-6.1 SolOpenAI |
|---|---|---|---|
| Weights | Open | Closed | Closed |
| Parameters | 2.8T / 104B active | — | — |
| Context | 1M | 1M | 1.05M |
| Precision published | fp4 · disclosed | no published weights to label | no published weights to label |
| Input / 1M | $1.75 | $2.00 | $2.00 |
| Output / 1M | $10.00 | $10.00 | $10.00 |
| Cache read / 1M | $0.18 | $0.10 | $0.10 |
| Blended / 1M3:1 input to output | $3.813 | $4.0001.0x | $4.0001.0x |
| Links | Model page | Vendor pricing | Vendor pricing |
First-party list rates verified Nov 13, 2026, each column links to its source
Blended is 0.75 × input + 0.25 × output — a mix for ranking, not a rate. Cache reads are excluded from it.
