# Browser LLM Bench

> Measured browser LLM inference performance and scoped resource usage. Curated public reports and a private, on-device runner.

Measurements describe an exact model artifact, quantization, runtime, browser, device, workload and cache state. Performance medians exclude warmup. Tokenizers and actual output lengths differ across models. These tests do not measure intelligence or task quality; no overall model score is assigned. Missing telemetry is unavailable, never zero. CPU is dedicated Chrome process-tree capacity; Chrome GPU time is experimental driver-accounted time including rendering, not GPU utilization. Power is whole-host subsystem power, not model-only power. Browser memory is summed process footprint, not deduplicated RAM.

The primary comparison loads one model once and runs one warmup followed by 60 measured requests in three windows of 20. Each request starts immediately after the previous response completes. There are 90-second loaded-idle gaps after windows one and two. After the third window: 30 seconds loaded idle, unload, then 15 seconds observation. Conversation history resets between requests. The windows share one session and are not independent trials. Older reports can use different schedules: read each report's saved protocol before comparing.

## Guides

- [Leaderboard and measured comparison](https://browserllm.dev/seo/guides/index.md): Published results and comparison limitations.
- [Methodology](https://browserllm.dev/seo/guides/methodology.md): Protocol, timing formulas, resource accounting and scoring status.
- [Private benchmark runner](https://browserllm.dev/seo/guides/run.md): Supported configuration options and local report behavior.
- [Compatibility](https://browserllm.dev/seo/guides/compatibility.md): Implemented runtimes, requirements and evidence states.

## Models

- [SmolLM2 135M](https://browserllm.dev/seo/guides/models/smollm2-135m.md): 0.135B parameters; published measurements available.
- [Qwen3 0.6B](https://browserllm.dev/seo/guides/models/qwen3-0-6b.md): 0.6B parameters; published measurements available.
- [Qwen3 4B](https://browserllm.dev/seo/guides/models/qwen3-4b.md): 4B parameters; published measurements available.
- [Qwen2.5 7B](https://browserllm.dev/seo/guides/models/qwen2-5-7b.md): 7B parameters; published measurements available.
- [Qwen2.5 0.5B](https://browserllm.dev/seo/guides/models/qwen2-5-0-5b.md): 0.5B parameters; published measurements available.
- [Qwen2.5 1.5B](https://browserllm.dev/seo/guides/models/qwen2-5-1-5b.md): 1.5B parameters; published measurements available.
- [Llama 3.2 1B](https://browserllm.dev/seo/guides/models/llama-3-2-1b.md): 1B parameters; published measurements available.
- [Qwen2.5 3B](https://browserllm.dev/seo/guides/models/qwen2-5-3b.md): 3B parameters; published measurements available.
- [Qwen3 1.7B](https://browserllm.dev/seo/guides/models/qwen3-1-7b.md): 1.7B parameters; published measurements available.
- [Mistral 7B Instruct v0.3](https://browserllm.dev/seo/guides/models/mistral-7b-v0-3.md): 7B parameters; published measurements available.
- [Mistral 7B Instruct v0.2](https://browserllm.dev/seo/guides/models/mistral-7b-v0-2.md): 7B parameters; published measurements available.
- [SmolLM2 1.7B](https://browserllm.dev/seo/guides/models/smollm2-1-7b.md): 1.7B parameters; configuration only, no published measurement.
- [Gemma 2 2B](https://browserllm.dev/seo/guides/models/gemma-2-2b.md): 2B parameters; configuration only, no published measurement.
- [Phi-4 Mini](https://browserllm.dev/seo/guides/models/phi-4-mini.md): 3.83602176B parameters; configuration only, no published measurement.
- [Gemma 4 E2B](https://browserllm.dev/seo/guides/models/gemma-4-e2b.md): 5.123178051B parameters; configuration only, no published measurement.
- [Qwen3.5 0.8B](https://browserllm.dev/seo/guides/models/qwen3-5-0-8b.md): 0.873438784B parameters; configuration only, no published measurement.
- [Qwen3.5 2B](https://browserllm.dev/seo/guides/models/qwen3-5-2b.md): 2.274069824B parameters; configuration only, no published measurement.
- [Qwen3.5 4B](https://browserllm.dev/seo/guides/models/qwen3-5-4b.md): 4.659865088B parameters; configuration only, no published measurement.
- [Ministral 3 3B](https://browserllm.dev/seo/guides/models/ministral-3-3b.md): 4.251743232B parameters; configuration only, no published measurement.

## Optional

- [Published report manifest](https://browserllm.dev/reports/manifest.json): Curated raw JSON evidence only; never private user runs.
- [Source repository](https://github.com/gufranmirza/browser-llm-bench): Static website, versioned catalog and local macOS harness.