PixCode models
Three aliases, 31 official names. What changes is what runs underneath; what stays is each alias’s role and its rate.
pixcode-fast
500turns / 1,000 crystals
Cheap enough to use freely
Built for an agent’s chores. Set it as small_model in opencode and titles and compaction become nearly free. It struggles alone on complex work — it is not meant to be the main driver.
Specs
- Context window 1M
- Input $0.14 / Output $0.28
- Available on Starter / Pro / Max
pixcode-proeveryday driver
175turns / 1,000 crystals
The one to put in your config
What you should be on most of the time. A 1M context means feeding in large parts of the codebase directly instead of retrieving in circles, and with prefix caching the repeated context costs very little.
Specs
- Context window 1M
- Input $0.43 / Output $0.87
- Available on Starter / Pro / Max
pixcode-max
25turns / 1,000 crystals
Finishes the long runs
The difference is not the benchmark score, it is whether the run finishes. In public evaluations, models that score alike diverge sharply past twenty steps — some time out and give up, some reach the end. This tier buys the latter.
Specs
- Context window 1M
- Input $1.40 / Output $4.40
- Available on Pro / Max
Turn counts are derived from the published rates, assuming a typical agent request of 30k input and 2k output.
Relative cost
Turns per the same amount of crystals. One max turn costs twenty fast turns, so do not give it the chores.
All models. Call the official names directly. The price is the upstream official list price — zero markup.
31 available · live rates
| Model | Context | API | Cache read | Input | Output |
|---|---|---|---|---|---|
| PixCode | |||||
| pixcode-max | 1M | anthropic | $0.26 | $1.40 | $4.40 |
| pixcode-fast | 1M | anthropic | $0.0032 | $0.16 | $0.32 |
| pixcode-pro | 1M | anthropic | $0.004 | $0.5 | $1.00 |
| Claude | |||||
| claude-fable-5 | 1M | anthropic | $1.00 | $10.00 | $50.00 |
| claude-opus-5 | 1M | anthropic | $0.5 | $5.00 | $25.00 |
| claude-sonnet-5 | 1M | anthropic | $0.2 | $2.00 | $10.00 |
| claude-haiku-4-5-20251001 | 200K | anthropic | $0.1 | $1.00 | $5.00 |
| GPT | |||||
| gpt-5.6-luna | 1.1M | openai | $0.02 | $0.2 | $1.20 |
| gpt-5.6-terra | 1.1M | openai | $0.2 | $2.00 | $12.00 |
| gpt-5.6-sol | 1.1M | openai | $0.5 | $5.00 | $30.00 |
| Gemini | |||||
| gemini-3.5-flash | 1.0M | openai | $0.15 | $1.50 | $9.00 |
| gemini-3.1-pro-preview | 1.0M | openai | $0.2 | $2.00 | $12.00 |
| DeepSeek | |||||
| deepseek-v3.2 | 164K | anthropic | $0.1345 | $0.269 | $0.4 |
| deepseek-v3.2-exp | 164K | anthropic | $0.27 | $0.27 | $0.41 |
| deepseek-v3.1 | 131K | anthropic | $0.135 | $0.27 | $1.00 |
| deepseek-v3-0324 | 164K | anthropic | $0.135 | $0.27 | $1.12 |
| deepseek-v3-turbo | 64K | anthropic | $0.4 | $0.4 | $1.30 |
| deepseek-r1-0528 | 164K | anthropic | $0.35 | $0.7 | $2.50 |
| deepseek-r1-turbo | 64K | anthropic | $0.7 | $0.7 | $2.50 |
| Kimi | |||||
| kimi-k3 | 1.0M | anthropic | $0.3 | $3.00 | $15.00 |
| kimi-k2.5 | 262K | anthropic | $0.1 | $0.6 | $3.00 |
| kimi-k2-thinking | 262K | anthropic | $0.15 | $0.6 | $2.50 |
| kimi-k2-0905 | 262K | anthropic | $0.6 | $0.6 | $2.50 |
| kimi-k2-instruct | 131K | anthropic | $0.57 | $0.57 | $2.30 |
| GLM | |||||
| glm-5.2 | 1.0M | anthropic | $0.26 | $1.40 | $4.40 |
| glm-5 | 203K | anthropic | $0.2 | $1.00 | $3.20 |
| glm-4.7 | 205K | anthropic | $0.11 | $0.6 | $2.20 |
| glm-4.6 | 205K | anthropic | $0.11 | $0.55 | $2.20 |
| glm-4.6v | 131K | anthropic | $0.055 | $0.3 | $0.9 |
| glm-4.5 | 131K | anthropic | $0.11 | $0.6 | $2.20 |
| glm-4.5v | 66K | anthropic | $0.11 | $0.6 | $1.80 |
USD per 1M tokens. Models marked anthropic also accept the OpenAI endpoint; openai-only models must use /v1/chat/completions. Cache-write rates are in the docs.
Start on pro, switch when you need to.
The aliases are three defaults. All 31 official names can be called by name at any time — one config line apart.