Use case

Best local LLM for customer support

This page does not name a winner. The title is the job. You test the suggested models on your own prompts, then decide.

The catalog has no customer-support tag. This page routes the job as chat and support.

Catalog label: chat and support. A catalog tag is a routing label, not a benchmark, and this site has no measured score for this job.

Test this use case

Example prompts

• A customer writes: my order arrived damaged and I need a replacement this week. Write a short reply that asks for the order number and does not promise a refund.
• Turn this rule into a support answer: refunds within 14 days if the item is unused. The customer says they opened the box.
• Summarise this ticket in three bullets an agent can act on: login fails after a password reset, the customer is on a phone, and they need access before 15:00.

The compare page can drop one of these in with one click. They are starters, not a test set we scored.

Open models that carry this tag

Alphabetical. Order here is not a ranking. The compare suggester reorders them from your languages, context, privacy, and budget, and it says why.

• DeepSeek V3Catalog tags: chat, code, reasoning. Routing labels, not benchmarks.
• Gemma 3 12BCatalog tags: chat, translation, creative. Routing labels, not benchmarks.
• Gemma 3 4BCatalog tags: chat, translation. Routing labels, not benchmarks.
• Gemma 4 26B A4BCatalog tags: chat. Routing labels, not benchmarks.
• Gemma 4 31BCatalog tags: chat. Routing labels, not benchmarks.
• GLM 4.7 FlashCatalog tags: chat, code, reasoning. Routing labels, not benchmarks.
• gpt-oss-120bCatalog tags: chat, code, reasoning. Routing labels, not benchmarks.
• gpt-oss-20bCatalog tags: chat, code, reasoning. Routing labels, not benchmarks.
• Kimi K2.6Catalog tags: chat, code, reasoning. Routing labels, not benchmarks.
• LFM2.5 2.6BCatalog tags: chat, reasoning, rag. Routing labels, not benchmarks.
• LFM2.5 24B A2BCatalog tags: chat, reasoning. Routing labels, not benchmarks.
• Llama 3.1 8B InstructCatalog tags: chat, code, rag. Routing labels, not benchmarks.
• Llama 3.2 1B InstructCatalog tags: chat. Routing labels, not benchmarks.
• Llama 3.2 3B InstructCatalog tags: chat, creative. Routing labels, not benchmarks.
• Llama 3.3 70B InstructCatalog tags: chat, rag, reasoning, code. Routing labels, not benchmarks.
• Llama 4 ScoutCatalog tags: chat. Routing labels, not benchmarks.
• Ministral 3 14BCatalog tags: chat, translation. Routing labels, not benchmarks.
• Ministral 3 3BCatalog tags: chat, translation. Routing labels, not benchmarks.
• Mistral Small 3.2 24BCatalog tags: chat, translation, rag, code. Routing labels, not benchmarks.
• Mistral Small 4Catalog tags: chat, code, reasoning, translation. Routing labels, not benchmarks.
• Nemotron 3 Nano 30B A3BCatalog tags: chat, reasoning. Routing labels, not benchmarks.
• Nemotron 3 Super 120BCatalog tags: chat, reasoning. Routing labels, not benchmarks.
• Qwen3 235B A22BCatalog tags: chat, reasoning, translation. Routing labels, not benchmarks.
• Qwen3.5 9BCatalog tags: chat, code, reasoning. Routing labels, not benchmarks.
• Qwen3.6 35B A3BCatalog tags: chat, code, reasoning. Routing labels, not benchmarks.

The compare also keeps one frontier reference, Claude Sonnet 5.5. It is not open weights, and it is not a hardware recommendation.

Size hardware after you know the model · Hardware catalog · All use cases