BenchLM

Ask an agent to

• Find the best models for coding
• Compare two AI models side by side
• Estimate monthly API cost for 10,000 blog posts

10 tools

When we loaded benchlm.ai again on Oct 24, 2026, it registered these through navigator.modelContext or provideContext(), from earlier WebMCP drafts. Chrome 154 only offers document.modelContext, so Chrome’s own agent support does not see them.

• compare_ai_models

  Return a BenchLM side-by-side comparison summary for two AI models, then open the comparison page in the current tab. Use this when a user wants benchmark, pricing, context window, or capability tradeoffs between…

  modelA: string, modelB: string

• estimate_ai_costread-only

  Estimate monthly, daily, and yearly AI model cost for real workloads such as blog posts, website pages, docs articles, PRDs/specs, or shipped product features.

  workload: string, unitsPerMonth: number?, wordsPerUnit: number?, revisionRounds: number?, complexity: string?, modelNames: array?

• find_best_modelsread-only

  Return top BenchLM-ranked AI models for a use case such as coding, reasoning, tool use, web research, computer use, long context, document AI, image understanding, factuality, open source, or value.

  useCase: string, limit: number?, openWeight: boolean?, maxOutputPrice: number?, minContextTokens: number?, reasoning: boolean?

• get_model_detailsread-only

  Return BenchLM metadata for one AI model, including provider, score, source type, context window, reasoning mode, pricing, speed, and category scores.

  model: string

• lookup_benchmark_resultsread-only

  Return the BenchLM leaderboard for a specific benchmark, such as SWE-bench Verified, LiveCodeBench, MMLU, GPQA, BFCL, BrowseComp, OSWorld, or AIME.

  benchmark: string, limit: number?, includeUnverified: boolean?

• open_benchlm_page

  Navigate the current BenchLM tab to the best matching page for a model, benchmark, provider, ranking, comparison, blog post, or tool query.

  query: string, type: string?

• open_model_profile

  Navigate to the BenchLM profile page for a specific AI model by model name, alias, or slug.

  model: string

• recommend_llmread-only

  Recommend AI models using BenchLM selector logic for a user use case, budget, context-window need, open-weight preference, and speed/quality preference.

  useCase: string, task: string?, audience: string?, budget: string?, context: string?, openWeight: string?, speed: string?, limit: number?

• search_benchlmread-only

  Search BenchLM for AI models, benchmarks, rankings, comparisons, tools, providers, and blog posts. Use this before navigating when the exact BenchLM page is unknown.

  query: string, limit: number?

• subscribe_to_benchlm_updates

  Subscribe an email address to weekly BenchLM benchmark, ranking, and pricing updates. This tool always asks the user to confirm before submitting.

  email: string