Voyage AI / voyageai/voyage-code-4
Voyage Code 4 - access through LLMTR
Voyage Code 4 succeeds Voyage Code 3 and is trained specifically for the retrieval coding agents need. Its training corpus is mined from closed pull requests, so the model learns to start from a vague symptom such as a bug report and locate the file that needs to change - exactly the step where grep's "what's broken" queries fall short. It supports 256, 512, 1024 and 2048 dimensional output (Matryoshka learning) alongside float32, int8, uint8 and binary quantization; the 32,000-token window covers most source files in a single pass. It costs a third less per million tokens than Voyage Code 3.
Technical specifications
| Canonical ID | voyageai/voyage-code-4 |
|---|---|
| Provider | Voyage AI |
| Context window | 32,000 tokens |
| Operations | EMBEDDINGS |
| Modalities | text, embedding |
Pricing
An 8% platform margin applies to credit top-ups; model usage prices are not separately marked up.
| Operation | Metric | Unit | Price |
|---|---|---|---|
| EMBEDDINGS | INPUT_TEXT | PER_1M_TOKENS | $0.120000 |
Example usage
With existing OpenAI SDK flows, change only the base URL and model identifier.
curl https://llmtr.com/v1/chat/completions \
-H "Authorization: Bearer llmtr-your_key" \
-H "Content-Type: application/json" \
-d '{"model":"voyageai/voyage-code-4","messages":[{"role":"user","content":"Hello"}]}'Related models
- Voyage 4 Large voyageai/voyage-4-large
- Voyage 4 voyageai/voyage-4
- Voyage Code 3 voyageai/voyage-code-3
- Voyage Context 4 voyageai/voyage-context-4
- Voyage Multimodal 3.5 voyageai/voyage-multimodal-3.5
- Voyage Rerank 2.5 voyageai/rerank-2.5
- Voyage 4 Lite voyageai/voyage-4-lite
- Voyage Finance 2 voyageai/voyage-finance-2