MODEL PRICING
Every token, accounted for.
Input and output rates, per million tokens.
Obit’s published list price, beside the market’s same-model average.
Qwen 3.8 27B is ready to use in the dashboard. Contact us to set up any of the other models.
QWENReady in the dashboard
Qwen 3.8 27B
USD / 1 million tokens
OBIT63–68% less
- Input
- $0.10
- Output
- $1.00
Cached input: $0.01 / 1M
Save $0.2146 per 1M input tokens and $1.6667 per 1M output tokens.
Compare intelligence & price
Artificial Analysis Intelligence Index v4.3. Higher scores are better. Prices are USD per 1M tokens.
Qwen 3.8 27B on Obit costs less per token than GPT-5.6 Luna and Claude Sonnet 5, on input and on output.
Nearby scores do not imply identical performance. Sources & methodology ↓
OPENAISetup by request
GPT-OSS 20B
USD / 1 million tokens
OBIT30% less
- Input
- $0.0293
- Output
- $0.1152
Cached input: $0.0029 / 1M
Save $0.0125 per 1M input tokens and $0.0494 per 1M output tokens.
Compare intelligence & price
Artificial Analysis Intelligence Index v4.3. Higher scores are better. Prices are USD per 1M tokens.
Models with nearby index scores| Model / effort | Index | Input | Output |
|---|
| GPT-OSS 20Bhigh · on Obit | 9 | $0.0293 | $0.1152 |
|---|
| GPT-4.1 mininon-reasoning | 10 est. | $0.40 | $1.60 |
|---|
Nearby scores do not imply identical performance. “est.” marks an Artificial Analysis estimate. Sources & methodology ↓
QWENSetup by request
Qwen 3.6 35B-A3B
USD / 1 million tokens
OBIT30% less
- Input
- $0.0961
- Output
- $0.7143
Cached input: $0.0096 / 1M
Save $0.0412 per 1M input tokens and $0.3061 per 1M output tokens.
Compare intelligence & price
Artificial Analysis Intelligence Index v4.3. Higher scores are better. Prices are USD per 1M tokens.
Models with nearby index scores| Model / effort | Index | Input | Output |
|---|
| Qwen 3.6 35B-A3Breasoning · on Obit | 19 | $0.0961 | $0.7143 |
|---|
| GPT-5 minihigh | 17 | $0.25 | $2.00 |
|---|
Nearby scores do not imply identical performance. Sources & methodology ↓
GOOGLESetup by request
Gemma 4 31B
USD / 1 million tokens
OBIT30% less
- Input
- $0.1925
- Output
- $0.4282
Cached input: $0.0193 / 1M
Save $0.0825 per 1M input tokens and $0.1835 per 1M output tokens.
Compare intelligence & price
Artificial Analysis Intelligence Index v4.3. Higher scores are better. Prices are USD per 1M tokens.
Models with nearby index scores| Model / effort | Index | Input | Output |
|---|
| Gemma 4 31Breasoning · on Obit | 15 | $0.1925 | $0.4282 |
|---|
| Claude Haiku 4.5non-reasoning | 15 est. | $1.00 | $5.00 |
|---|
Nearby scores do not imply identical performance. “est.” marks an Artificial Analysis estimate. Sources & methodology ↓
GOOGLESetup by request
Gemma 4 26B-A4B
USD / 1 million tokens
OBIT30% less
- Input
- $0.0727
- Output
- $0.2565
Cached input: $0.0073 / 1M
Save $0.0311 per 1M input tokens and $0.1099 per 1M output tokens.
Compare intelligence & price
Artificial Analysis Intelligence Index v4.3. Higher scores are better. Prices are USD per 1M tokens.
Models with nearby index scores| Model / effort | Index | Input | Output |
|---|
| Gemma 4 26B-A4Breasoning · on Obit | 17 est. | $0.0727 | $0.2565 |
|---|
| GPT-5 minihigh | 17 | $0.25 | $2.00 |
|---|
Nearby scores do not imply identical performance. “est.” marks an Artificial Analysis estimate. Sources & methodology ↓
Pricing sources & comparison methodology · market snapshot September 13, 2026
Obit’s rates are list prices, not a discount off anyone else’s. Every live model’s input, output, and cached-input price is declared in pricing.json — the machine-readable source for the OpenAI-compatible API at https://relay.obitmc.com/v1. This page renders that declaration and re-reads the same file in your browser, so a change to a list price shows here without a site deploy. Cached input is a prompt prefix served from a worker’s prefix cache rather than recomputed, so a repeated prefix costs less. It is priced 90% below that model’s input rate. The first send of a prefix pays the full rate.
The market average is the arithmetic mean of available paid providers for the same model on OpenRouter, captured September 13, 2026. We average each provider’s available endpoints first, then weight providers equally. Free and unavailable endpoints are excluded. Input and output are compared separately, and every percentage on this page is (market − Obit) ÷ market for that model, rounded to a whole percent.
Models marked “setup by request” are not live on Obit yet and have no declared list price. Their figures are indicative, 30% below that same market average, and their cached rate sits 90% below their input rate. We confirm the rate when we set one up.
Market comparisons use standard uncached text-token rates in USD per million tokens. Batch discounts, other modalities, and custom cloud deployments are excluded. Total workload cost depends on token usage, including reasoning tokens.
Comparisons use the same Artificial Analysis Intelligence Index v4.3, with the evaluated reasoning effort shown for each model. We selected GPT/Claude models within four rounded index points; scores describe the evaluated models, not a benchmark of Obit’s deployments. Some reference models are older generations.