InternLM2-Math Models by Intern-AI
About
The InternLM2-Math family is a collection of advanced, bilingual (Chinese and English) open-source large language models specifically tuned for mathematical reasoning. These models act as solvers, provers, verifiers, and augmentors, excelling at both formal and informal mathematical reasoning tasks. They are pretrained on around 100 billion quality math-related tokens and further refined with about 2 million bilingual math-supervised data points. Notably, the models support Lean 3, a formal proof assistant that enables verifiable mathematical reasoning. Available in different variants with sizes like 7B and 20B parameters, these models offer varying levels of performance on multiple benchmarks. The InternLM2-Math-Plus series features enhancements for improved performance in both formal and informal reasoning tasks, epitomizing the latest advancements in this LLM family.
Current Variants
Use-when guidance is derived from seed capabilities, context, release, and replacement fields.
Use when the workload needs 4k context and 20B parameters.
Use when the workload needs 4k context and 7B parameters.
Use when the workload needs 4k context and 20B parameters.
Use when the workload needs 4k context and 7B parameters.
Use when the workload needs 4k context and 1.8B parameters.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| InternLM2 Math 20B | Use when the workload needs 4k context and 20B parameters. | 2024-05 | 4k context20B parameters | Current |
| InternLM2 Math 7B | Use when the workload needs 4k context and 7B parameters. | 2024-05 | 4k context7B parameters | Current |
| InternLM2 Math Plus 20B | Use when the workload needs 4k context and 20B parameters. | 2024-05 | 4k context20B parameters | Current |
| InternLM2 Math Plus 7B | Use when the workload needs 4k context and 7B parameters. | 2024-05 | 4k context7B parameters | Current |
| InternLM2 Math Plus 1.8B | Use when the workload needs 4k context and 1.8B parameters. | 2024-05 | 4k context1.8B parameters | Current |
| InternLM2 Math Plus Mixtral 8x22B | Use when the workload needs 64k context. | 2024-05 | 64k context | Current |
Release Timeline
1 release groupSpecifications(6 models)
| Model | Released | Context | Parameters |
|---|---|---|---|
| InternLM2 Math 20B | 2024-05 | 4k | 20B |
| InternLM2 Math 7B | 2024-05 | 4k | 7B |
| InternLM2 Math Plus 20B | 2024-05 | 4k | 20B |
| InternLM2 Math Plus 7B | 2024-05 | 4k | 7B |
| InternLM2 Math Plus 1.8B | 2024-05 | 4k | 1.8B |
| InternLM2 Math Plus Mixtral 8x22B | 2024-05 | 64k | 8x22B |
Frequently Asked Questions
- What is InternLM2-Math used for?
- InternLM2-Math is used for mathematics, coding, and math-heavy prompts. The family description and listed model capabilities point to those workloads as the best fit.
- How does InternLM2-Math compare to InternVL?
- InternLM2-Math by Intern-AI is strongest where you need mathematics, while InternVL by Intern-AI is the closest related family to check for adjacent model selection. InternLM2-Math has 6 listed variants and reaches up to 64k context, while InternVL reaches up to 32k context, so compare the specs and pricing tables before choosing a production model.
- Which InternLM2-Math model should I use?
- If price is the main constraint, use the pricing table first because InternLM2-Math does not have complete provider pricing in the local data. For the most capable/latest local choice, evaluate InternLM2 Math Plus Mixtral 8x22B with 64k context.
