metrillm

分类: 数据与AI | 上传者: MetriLLM | 下载: 0 | 版本: v1.0（最新）

Find the best local LLM for your machine. Tests speed, quality and RAM fit, then tells you if a model is worth running on your hardware.

更新日志: Source: GitHub https://github.com/MetriLLM/metrillm

目录结构

当前层级: 根目录

📁 .github/
- 📁 ISSUE_TEMPLATE/
  - 📄 bug_report.yml 1.9 KB
  - 📄 config.yml 253 B
  - 📄 feature_request.yml 1.6 KB
- 📁 workflows/
  - 📄 ci.yml 2.8 KB
  - 📄 release.yml 5.4 KB
- 📄 branch-protection.md 1.0 KB
- 📄 FUNDING.yml 19 B
- 📄 PULL_REQUEST_TEMPLATE.md 591 B
📁 docs/
- 📁 images/
  - 📄 cli1.png 964.6 KB
  - 📄 cli2.png 727.9 KB
  - 📄 leaderboard.png 4.0 MB
  - 📄 social-preview.png 283.0 KB
- 📄 multi-backend.md 4.7 KB
📁 Formula/
- 📄 metrillm.rb 679 B
📁 mcp/
- 📁 src/
  - 📄 index.ts 3.4 KB
  - 📄 setup.ts 526 B
  - 📄 tools.ts 9.1 KB
- 📄 package-lock.json 105.5 KB
- 📄 package.json 1.3 KB
- 📄 README.md 3.5 KB
- 📄 server.json 604 B
- 📄 tsconfig.json 480 B
- 📄 tsup.config.ts 601 B
📁 plugins/
- 📁 claude-code/
  - 📁 .claude-plugin/
    
    📄 plugin.json 450 B
  - 📁 agents/
    
    📄 benchmark-advisor.md 2.4 KB
  - 📁 skills/
    
    📁 benchmark/
    
    📄 SKILL.md 3.3 KB
    
    📁 metrillm-guide/
    
    📄 SKILL.md 1.9 KB
  - 📄 .mcp.json 105 B
  - 📄 README.md 1.7 KB
- 📁 cursor/
  - 📁 .cursor-plugin/
    
    📄 plugin.json 512 B
  - 📁 agents/
    
    📄 benchmark-advisor.md 2.4 KB
  - 📁 rules/
    
    📄 metrillm-model-selection.mdc 1.3 KB
  - 📁 skills/
    
    📁 benchmark/
    
    📄 SKILL.md 3.3 KB
    
    📁 metrillm-guide/
    
    📄 SKILL.md 1.9 KB
  - 📄 .mcp.json 105 B
  - 📄 README.md 2.0 KB
📁 prd/
- 📁 evolutions/
  - 📄 prd-tty-migration.md 14.0 KB
- 📄 README.md 208 B
📁 scripts/
- 📄 bench-mlx-overnight.sh 7.7 KB
- 📄 e2e-smoke.ts 4.0 KB
- 📄 list-lmstudio-rows.ts 874 B
- 📄 rescore-db.ts 3.7 KB
- 📄 update-homebrew-formula.sh 1.5 KB
📁 src/
- 📁 benchmarks/
  - 📄 coding.ts 14.4 KB
  - 📄 instruction-following.ts 8.7 KB
  - 📄 math.ts 2.6 KB
  - 📄 multilingual.ts 5.0 KB
  - 📄 performance.ts 12.0 KB
  - 📄 profile.ts 1006 B
  - 📄 reasoning.ts 2.4 KB
  - 📄 structured-output.ts 12.1 KB
- 📁 commands/
  - 📄 bench.ts 18.1 KB
  - 📄 list.ts 3.5 KB
- 📁 core/
  - 📄 app-meta.ts 2.1 KB
  - 📄 exporter.ts 7.2 KB
  - 📄 hardware.ts 7.9 KB
  - 📄 lm-studio-client.ts 45.8 KB
  - 📄 ollama-client.ts 7.9 KB
  - 📄 runtime-unavailable.ts 563 B
  - 📄 runtime.ts 7.2 KB
  - 📄 store.ts 3.5 KB
  - 📄 submitter.ts 1.4 KB
  - 📄 telemetry.ts 2.8 KB
  - 📄 update-checker.ts 4.2 KB
  - 📄 uploader.ts 7.4 KB
- 📁 datasets/
  - 📄 coding.json 56.6 KB
  - 📄 instruction-following.json 4.0 KB
  - 📄 math.json 6.0 KB
  - 📄 multilingual.json 3.9 KB
  - 📄 reasoning.json 11.9 KB
  - 📄 structured-output.json 4.6 KB
- 📁 scoring/
  - 📄 fitness.ts 8.5 KB
  - 📄 performance-scorer.ts 5.7 KB
  - 📄 quality-scorer.ts 2.7 KB
- 📁 ui/
  - 📄 banner.ts 3.9 KB
  - 📄 guru-meditation.ts 3.8 KB
  - 📄 menu.ts 32.0 KB
  - 📄 progress.ts 4.7 KB
  - 📄 results-table.ts 12.3 KB
  - 📄 score-color.ts 269 B
  - 📄 share-prompt.ts 5.6 KB
  - 📄 submitter-prompt.ts 5.4 KB
  - 📄 terminal.ts 327 B
  - 📄 thinking-prompt.ts 1.1 KB
  - 📄 verdict.ts 8.7 KB
- 📄 cli-interactive.ts 170 B
- 📄 cli-main.ts 12.8 KB
- 📄 index.ts 1.2 KB
- 📄 types.ts 7.6 KB
- 📄 utils.ts 19.8 KB
📁 supabase/
- 📁 migrations/
  - 📄 20260228163418_create_benchmarks_table.sql 1.9 KB
  - 📄 20260228190000_add_model_info_columns.sql 826 B
  - 📄 20260228230000_create_benchmark_leads_table.sql 896 B
  - 📄 20260228234500_add_benchmark_leads_update_policy.sql 535 B
  - 📄 20260301150000_add_hardware_and_thinking_columns.sql 1.1 KB
  - 📄 20260301160000_backfill_machine_model_jsonb.sql 493 B
  - 📄 20260301170000_add_get_rank_function.sql 1.3 KB
  - 📄 20260301180000_add_runtime_backend_column.sql 310 B
  - 📄 20260303100000_reweight_global_score_30_70.sql 1.4 KB
  - 📄 20260303110000_delete_coding_bug_rows.sql 1.6 KB
  - 📄 20260305133000_backfill_hardware_normalization.sql 2.5 KB
  - 📄 20260305134500_fix_hardware_backfill_regex.sql 1.6 KB
  - 📄 20260305141000_backfill_ryzen_ai_max_gpu.sql 410 B
  - 📄 20260305230500_allow_null_memory_percent_when_footprint_unavailable.sql 220 B
📁 tests/
- 📄 bench-share-policy.test.ts 12.6 KB
- 📄 benchmark-validators.test.ts 15.6 KB
- 📄 cli-interactive.test.ts 742 B
- 📄 coding-eval.test.ts 15.0 KB
- 📄 coding-sandbox-regression.test.ts 2.5 KB
- 📄 datasets.test.ts 6.3 KB
- 📄 exporter.test.ts 9.3 KB
- 📄 hardware.test.ts 17.1 KB
- 📄 list-command-runtime-selection.test.ts 2.1 KB
- 📄 list-command.test.ts 3.6 KB
- 📄 lm-studio-client.test.ts 30.4 KB
- 📄 mcp-tools-get-results.test.ts 2.9 KB
- 📄 mcp-tools-runtime.test.ts 3.9 KB
- 📄 menu-flow.test.ts 1.9 KB
- 📄 menu-rendering.test.ts 4.3 KB
- 📄 menu-settings.test.ts 6.5 KB
- 📄 ollama-client.test.ts 2.9 KB
- 📄 open-url.test.ts 1.5 KB
- 📄 performance-bench.test.ts 16.2 KB
- 📄 progress.test.ts 1.2 KB
- 📄 README.md 2.1 KB
- 📄 runtime-unavailable.test.ts 1.2 KB
- 📄 runtime.test.ts 6.6 KB
- 📄 sandbox.test.ts 3.9 KB
- 📄 scoring.test.ts 30.8 KB
- 📄 share-prompt.test.ts 7.0 KB
- 📄 store.test.ts 6.2 KB
- 📄 submitter-prompt.test.ts 2.5 KB
- 📄 telemetry.test.ts 3.6 KB
- 📄 ui.test.ts 16.5 KB
- 📄 update-checker.test.ts 7.3 KB
- 📄 uploader.test.ts 12.2 KB
- 📄 utils.test.ts 17.2 KB
📄 .env.example 592 B
📄 .gitignore 253 B
📄 .nvmrc 3 B
📄 AGENTS.md 4.8 KB
📄 CHANGELOG.md 3.8 KB
📄 CLAUDE.md 6.2 KB
📄 CODE_OF_CONDUCT.md 4.9 KB
📄 CONTRIBUTING.md 3.4 KB
📄 LICENSE 11.0 KB
📄 NOTICE 1.5 KB
📄 package-lock.json 93.0 KB
📄 package.json 1.6 KB
📄 README.md 11.0 KB
📄 SECURITY.md 1.7 KB
📄 SKILL.md 2.7 KB
📄 tsconfig.json 486 B
📄 tsup.config.ts 918 B
📄 vitest.config.ts 491 B

SKILL.md

---
name: metrillm
description: Find the best local LLM for your machine. Tests speed, quality and RAM fit, then tells you if a model is worth running on your hardware.
argument-hint: "[model-name]"
author: MetriLLM
source: https://github.com/MetriLLM/metrillm
license: Apache-2.0
allowed-tools: Bash, Read
install: npm install -g metrillm@latest
---

# MetriLLM — Find the Best LLM for Your Hardware

Test any local model and get a clear verdict: is it worth running on your machine?

## Prerequisites

1. **Node.js 20+** — check with `node -v`
2. **Ollama** or **LM Studio** installed and running
   - Ollama: [ollama.com](https://ollama.com), then `ollama serve`
   - LM Studio: [lmstudio.ai](https://lmstudio.ai), load a model and start the server
3. **MetriLLM CLI** — install globally:

```bash
npm install -g metrillm@latest
```

## Usage

### List available models

```bash
ollama list
```

### Run a full benchmark

```bash
metrillm bench --model $ARGUMENTS --json
```

This measures:
- **Performance**: tokens/second, time to first token, memory usage
- **Quality**: reasoning, math, coding, instruction following, structured output, multilingual
- **Fitness verdict**: EXCELLENT / GOOD / MARGINAL / NOT RECOMMENDED

### Performance-only benchmark (faster)

```bash
metrillm bench --model $ARGUMENTS --perf-only --json
```

Skips quality evaluation — measures speed and memory only.

### View previous results

```bash
ls ~/.metrillm/results/
```

Read any JSON file to see full benchmark details.

### Share to the public leaderboard

```bash
metrillm bench --model $ARGUMENTS --share
```

Uploads your result to the [MetriLLM community leaderboard](https://metrillm.dev) — an open, community-driven ranking of local LLM performance across real hardware. Compare your results with others and help the community find the best models for every setup. Shared data includes: model name, scores, hardware specs (CPU, RAM, GPU). No personal data is sent.

## Interpreting Results

| Verdict | Score | Meaning |
|---|---|---|
| EXCELLENT | >= 80 | Fast and accurate — great fit |
| GOOD | >= 60 | Solid — suitable for most tasks |
| MARGINAL | >= 40 | Usable but with tradeoffs |
| NOT RECOMMENDED | < 40 | Too slow or inaccurate |

Key metrics to highlight:
- `tokensPerSecond` > 30 = good for interactive use
- `ttft` < 500ms = responsive
- `memoryUsedGB` vs available RAM = will it fit?

## Tips

- Use `--perf-only` for quick tests
- Close GPU-intensive apps before benchmarking
- Benchmark duration varies depending on model speed and response length

## Open Source

MetriLLM is free and open source (Apache 2.0). Contributions, issues, and feedback are welcome: [github.com/MetriLLM/metrillm](https://github.com/MetriLLM/metrillm)

登录后下载/点赞/收藏 ❤ 3 | ★ 0

metrillm

目录结构

SKILL.md

举报内容

提示