> ## Documentation Index
> Fetch the complete documentation index at: https://docs.trajectory.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Models and pricing

> Per-token prices for training, evaluation, grading, and inference, plus compute and storage.

Prices are in USD per million tokens unless noted.

## Trainable models

Every trainable model can also be evaluated and deployed. Evaluation uses the same
rates, without the train charge.

| Model | Slug | Prefill | Cached prefill | Sample | Train |
| - | - | -: | -: | -: | -: |
| GLM 5.3 | `z-ai/glm-5.3` | \$3.6288 | \$0.4864 | \$11.664 | \$13.9968 |
| Qwen 3.5 397B | `Qwen/Qwen3.5-397B-A17B` | \$1.584 | \$0.2112 | \$7.20 | \$11.52 |
| Nemotron 3 Ultra | `nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16` | \$1.98 | \$0.288 | \$6.8475 | \$10.956 |
| Qwen 3.8 27B | `Qwen/Qwen3.8-27B` | \$0.672 | \$0.0896 | \$5.376 | \$7.1616 |
| GLM 5.3 Flash | `z-ai/glm-5.3-flash` | \$0.8064 | \$0.1088 | \$2.0256 | \$5.3568 |
| Inkling Small | `thinkingmachines/Inkling-Small` | \$0.638 | \$0.0928 | \$1.584 | \$3.817 |
| Nemotron 3.5 Super | `nvidia/NVIDIA-Nemotron-3.5-Super-120B-A12B-BF16` | \$0.627 | \$0.0912 | \$1.584 | \$2.552 |
| Nemotron 3 Super | `nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16` | \$0.627 | \$0.0912 | \$1.584 | \$2.552 |
| Qwen 3.6 35B | `Qwen/Qwen3.6-35B-A3B` | \$0.432 | \$0.0576 | \$1.2864 | \$2.0544 |
| Qwen 3.5 4B | `Qwen/Qwen3.5-4B` | \$0.3168 | \$0.0416 | \$0.9696 | \$1.2864 |
| Nemotron 3.5 Lightning | `nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16` | \$0.2024 | \$0.0288 | \$0.5445 | \$0.88 |

### Deployments

[Deployments](https://docs.trajectory.ai/guides/deployments.md) are billed per token at the base model's prefill,
cached prefill, and sample rates. There is no hourly charge.

## Evaluation models

These models can't be trained. Use them for evaluation and grading. GPT-6 Astra,
GPT-5.6 Sol, GPT-5.6 Luna, GPT-5.4 Mini, Claude Opus 5.5, and Claude Sonnet 5.5 are also on
the [inference API](https://docs.trajectory.ai/api-reference/inference/create-chat-completion-1.md). GPT-5.6 Terra is
available on the inference API, but not for evaluation or grading.

| Model | Slug | Prefill | Cached prefill | Cache write | Sample |
| - | - | -: | -: | -: | -: |
| GPT-6 Astra | `openai/gpt-6-astra` | \$13.00 | \$1.30 | \$16.25 | \$65.00 |
| GPT-5.5 | `openai/gpt-5.5` | \$6.50 | \$0.65 | \$6.50 | \$39.00 |
| GPT-5.6 Sol | `openai/gpt-5.6-sol` | \$5.20 | \$0.52 | \$6.50 | \$26.00 |
| Claude Opus 5.5 | `anthropic/claude-opus-5.5` | \$5.20 | \$0.26 | \$6.50 | \$26.00 |
| Claude Sonnet 4.6 | `anthropic/claude-sonnet-4.6` | \$3.90 | \$0.39 | \$4.875 | \$19.50 |
| Kimi K3 | `moonshotai/kimi-k3` | \$3.90 | \$0.39 | \$3.90 | \$19.50 |
| Gemini 3 Pro | `google/gemini-3.1-pro-preview` | \$2.60 | \$0.26 | \$0.4875 | \$15.60 |
| GPT-5.6 Terra | `openai/gpt-5.6-terra` | \$2.60 | \$0.26 | \$3.25 | \$15.60 |
| GPT-6.1 Sol | `openai/gpt-6.1-sol` | \$2.60 | \$0.13 | \$3.25 | \$13.00 |
| GPT-6 Sol | `openai/gpt-6-sol` | \$2.60 | \$0.26 | \$3.25 | \$13.00 |
| Claude Sonnet 5.5 | `anthropic/claude-sonnet-5.5` | \$2.60 | \$0.13 | \$3.25 | \$13.00 |
| GPT-5.4 Mini | `openai/gpt-5.4-mini` | \$0.975 | \$0.0975 | \$0.975 | \$5.85 |
| GPT-5 Mini | `openai/gpt-5-mini` | \$0.325 | \$0.0325 | \$0.325 | \$2.60 |
| GPT-5.6 Luna | `openai/gpt-5.6-luna` | \$0.26 | \$0.026 | \$0.325 | \$1.56 |

Claude cache writes are 5-minute; 1-hour writes cost 2× prefill.

### Long-context prompts

Requests whose prompt exceeds the threshold are billed entirely at these rates.

| Model | Slug | Prompt above | Prefill | Cached prefill | Cache write | Sample |
| - | - | -: | -: | -: | -: | -: |
| GPT-6 Astra | `openai/gpt-6-astra` | 272K tokens | \$26.00 | \$2.60 | \$32.50 | \$97.50 |
| GPT-5.5 | `openai/gpt-5.5` | 272K tokens | \$13.00 | \$1.30 | \$13.00 | \$58.50 |
| GPT-5.6 Sol | `openai/gpt-5.6-sol` | 272K tokens | \$10.40 | \$1.04 | \$13.00 | \$39.00 |
| Gemini 3 Pro | `google/gemini-3.1-pro-preview` | 200K tokens | \$5.20 | \$0.52 | \$0.4875 | \$23.40 |
| GPT-5.6 Terra | `openai/gpt-5.6-terra` | 272K tokens | \$5.20 | \$0.52 | \$6.50 | \$23.40 |
| GPT-6.1 Sol | `openai/gpt-6.1-sol` | 272K tokens | \$5.20 | \$0.26 | \$6.50 | \$19.50 |
| GPT-6 Sol | `openai/gpt-6-sol` | 272K tokens | \$5.20 | \$0.52 | \$6.50 | \$19.50 |
| GPT-5.6 Luna | `openai/gpt-5.6-luna` | 272K tokens | \$0.52 | \$0.052 | \$0.65 | \$2.34 |

## Grading

Model-based graders are billed at the grader model's token rates above.

## Sandbox environments

Each trajectory runs in its own sandbox, billed per second from ready to stop.

| Resource | Price |
| - | -: |
| CPU | \$0.20 per CPU-hour |
| Memory | \$0.04 per GiB-hour |

Set `cpus` and `memory_mb` in [`env_resources`](https://docs.trajectory.ai/guides/spec). Sandboxes can burst
above their request, so you can size below peak. Each second is billed at the
greater of the request and actual usage.

Example: 1 CPU and 4 GiB cost $0.36/hour, so 1,000 ten-minute trajectories cost $60.

## Trace storage

First 10 GB free each month, then \$0.15 per GB-month.

## Pricing terms

| Term | Meaning |
| - | - |
| Prefill | Prompt tokens the model processes. |
| Cached prefill | Prompt tokens served from the cache. |
| Cache write | Prompt tokens an evaluation model writes to its cache. |
| Sample | Tokens the model generates, including reasoning tokens. |
| Train | Tokens in a training step's forward and backward pass. |


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.