Phind Models by Phind
About
Phind's family of large language models excel in code generation, with enhancements over the open-source CodeLlama-34B foundation model. Notably, the Phind Model V7 achieves a HumanEval score of 74.7%, outperforming GPT-4 in coding tasks and operating at five times its speed by leveraging NVIDIA H100 GPUs and the TensorRT-LLM library for rapid processing of up to 100 tokens per second 12. With these advancements, Phind models support extensive context lengths of up to 16,000 tokens, prioritizing user input and web results. Furthermore, earlier versions like Phind-CodeLlama-34B-v2 are open-source on Hugging Face, allowing for independent capability assessments 310.
Current Variants
Use-when guidance is derived from seed capabilities, context, release, and replacement fields.
Use when the workload needs 32k context and 70B parameters.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Phind Instant | Use when the workload needs 32k context. | 2024-02 | 32k context | Current |
| Phind 70B | Use when the workload needs 32k context and 70B parameters. | 2024-02 | 32k context70B parameters | Current |
Release Timeline
1 release groupSpecifications(2 models)
| Model | Released | Context | Parameters |
|---|---|---|---|
| Phind Instant | 2024-02 | 32k | — |
| Phind 70B | 2024-02 | 32k | 70B |
Frequently Asked Questions
- What is Phind used for?
- Phind is used for coding. The family description and listed model capabilities point to those workloads as the best fit.
- How does Phind compare to Phind CodeLlama?
- Phind by Phind is strongest where you need coding, while Phind CodeLlama by Phind is the closest related family to check for coding. Phind has 2 listed variants and reaches up to 32k context, while Phind CodeLlama reaches up to 8k context, so compare the specs and pricing tables before choosing a production model.
- Which Phind model should I use?
- If price is the main constraint, use the pricing table first because Phind does not have complete provider pricing in the local data. For the most capable/latest local choice, evaluate Phind Instant with 32k context.
