A math explanation needs verification even when it sounds clear. This list proposes candidates for local evaluation. It provides no accuracy rank or iPhone performance result.
We checked the linked vendor sources on October 14, 2026. This post reports no comparative product test.
Check math solutions, units and assumptions
- Choose a problem with a known solution.
- Record the model, prompt, and any external tools.
- Check units and assumptions.
- Verify each calculation independently.
Qwen 2.5 Math 7B, Phi 3.5 Mini, Qwen 2.5 7B and Qwen 2.5 3B
1. Qwen2.5-Math-7B-Instruct
Qwen describes this as a math model. Its model card lists Apache 2.0. Check whether your setup uses the tools that a proposed task needs.
2. Phi-3.5-mini-instruct
Microsoft publishes this instruction model under MIT. See the model card. Compare each calculation with an independent solution.
3. Qwen2.5-7B-Instruct
Qwen publishes this general instruction model under Apache 2.0. See the model card. Use a known problem to check its explanation.
4. Qwen2.5-3B-Instruct
The Qwen model card lists the qwen-research license. Review those terms before use. This entry does not describe a current PocketLLM model.
PocketLLM is pre-launch. Private, on-device AI, starting on iPhone, with more platforms planned. No account required, no in-app tracking, and chat processing stays on your device. Join the waitlist.
Qwen 2.5 and Phi 3.5 Mini: math and license checks
| Option | Check before use |
|---|---|
| 1. Qwen2.5-Math-7B-Instruct | Check the final answer and tool use. |
| 2. Phi-3.5-mini-instruct | Check each calculation. |
| 3. Qwen2.5-7B-Instruct | Check the assumptions. |
| 4. Qwen2.5-3B-Instruct | Check the license and output. |
Before local use
Check the model file and runtime requirements on your device. A parameter count alone does not establish device compatibility.
Local inference keeps prompts on the device during inference when the selected model runs there. For a local-only workflow, select no cloud model. Disable chat sync and cloud backup. Check the app telemetry policy. Disable optional network features. Review extensions, exports, and local storage access.
Frequently asked questions
Which license does Qwen2.5-Math-7B-Instruct use?
The Qwen math model card lists Apache 2.0.
Which license does Phi-3.5-mini-instruct use?
Microsoft publishes this instruction model under MIT. See its model card.
Does Qwen2.5-3B-Instruct use the same license as Qwen2.5-7B-Instruct?
The Qwen 3B model card lists the qwen-research license. The Qwen model card lists Apache 2.0 for the 7B instruction model.
PocketLLM is pre-launch, with a waitlist for the iPhone launch. More platforms are planned. See the PocketLLM model catalog for its separate model list.