Your own plan
Connect your plans
Included
Provider-plan connections are included with your harness. You pay provider subscriptions separately.
- Anthropic · Claude
- OpenAI · ChatGPT / Codex
- xAI · Grok
Model rate card
What the models cost.
Bench runs on your own Claude, ChatGPT, or Grok plan. Here is what the labs charge.
Tokens are what the frontier models charge.
Where the thinking comes from
Connect any number or mix of your own AI provider plans to your harness at no extra charge. You pay provider subscriptions separately.
Enterprise
Enterprise plans can add Fuel. Your Bench rep sets it up.
Talk to your Bench repWhat a run costs at lab prices
Flagships · what the labs charge
Input and output rates per million tokens. Grok bills the whole request at a higher rate from 200,000 input tokens. OpenAI does the same from 272,000. Anthropic does not. The calculator uses those steps. Enterprise Fuel is priced in your agreement.
Anthropic
Claude Opus 5
- What the labs charge
- $5.00 / $25.00
OpenAIPromo at least through November 21, 2026
GPT-5.6 Sol
- What the labs charge
- $4.00 / $20.00
Promo at least through November 21, 2026. Standard rate $5.00 / $30.00.
xAI
Grok 4.7
- What the labs charge
- $2.00 / $6.00
Same job, lab prices
Pick a token shape to see what the labs charge for it.
- Input tokens
- 40,000
- Output tokens
- 6,000
- Cached share
- 20%
| Model | Labs charge |
|---|---|
| Claude Haiku 5.5 | $0.0063 |
| GPT-5.6 Luna | $0.01 |
| Grok 4.3 | $0.06 |
| Claude Haiku 4.5 | $0.06 |
| Grok 4.5 | $0.10 |
| Grok 4.7 | $0.10 |
| Grok 4.6 | $0.10 |
| Claude Sonnet 5 | $0.13 |
| GPT-5.6 Terra | $0.14 |
| GPT-5.6 Sol | $0.25 |
| Claude Opus 5 | $0.31 |
| Fable 5.1 | $0.62 |
| GPT-6 Astra | $0.63 |
The seats we run
| Seat | Model | Labs · in / out |
|---|---|---|
| Lead (unclear / T4) | Fable 5.1 | $10.00 / $50.00 |
| Lead (known shape) / panel | Claude Opus 5 | $5.00 / $25.00 |
| Pre-map / workers | Claude Sonnet 5 | $2.00 / $10.00 |
| Trivial | Claude Haiku 4.5 | $1.00 / $5.00 |
| Builder + first adversarial | GPT-6 Astra | $10.00 / $50.00 |
| Daybreak Blue | GPT-5.6 Sol | $4.00 / $20.00 |
| Parent | Grok 4.7 | $2.00 / $6.00 |
Full current lineup
| From | Model | Labs · in / out |
|---|---|---|
| Anthropic | Fable 5.1 | $10.00 / $50.00 |
| Anthropic | Claude Opus 5 | $5.00 / $25.00 |
| Anthropic | Claude Sonnet 5 | $2.00 / $10.00 |
| Anthropic | Claude Haiku 5.5 | $0.10 / $0.50 |
| Anthropic | Claude Haiku 4.5 | $1.00 / $5.00 |
| OpenAI | GPT-6 Astra | $10.00 / $50.00 |
| OpenAI | GPT-5.6 Sol | $4.00 / $20.00 |
| OpenAI | GPT-5.6 Terra | $2.00 / $12.00 |
| OpenAI | GPT-5.6 Luna | $0.20 / $1.20 |
| xAI | Grok 4.7 | $2.00 / $6.00 |
| xAI | Grok 4.6 | $2.00 / $6.00 |
| xAI | Grok 4.5 | $2.00 / $6.00 |
| xAI | Grok 4.3 | $1.25 / $2.50 |
Grok 4.1 Fast has no public token price from the labs, so it is not on this table.
Glossary
- Tokens
- What the frontier models charge for a run. Input is what the model reads. Output is what it writes.
- OAuth
- How you sign in to Claude, ChatGPT, or Grok so Bench can use the plan you already pay for. No password handed to us. Connect any number or mix of your own provider plans.
- Fuel
- Prepaid thinking Bench sells to enterprise customers through a rep. Most companies plug in their own plan.
- Harness
- The computer running your agents. The first harness is included with your first agent; each additional Gateway is $299 a month. Setup is $525 per harness. Nodes are $99 monthly network or storage devices without a harness or agents.
- Frontier labs
- xAI, Anthropic, and OpenAI. They make Grok, Claude, and ChatGPT / Codex.