Skip to main content
Glama

📊 LLM Latency Tracker

Measure AI API latency and uptime across regions, and publish the results as open data.

🌐 Live rankings · JSON API · MCP endpoint

⚡ What it does

  • 🌍 Regional measurements — compare provider response times and availability from distributed probe nodes.

  • ⏱️ Two probe types — network time to first byte (TTFB) and optional inference time to first token (TTFT).

  • 📈 Open dataset — daily rankings, historical snapshots and regional p50 / p95 statistics.

  • 🤖 Agent access — JSON API, MCP, Markdown pages and an llms.txt index.

  • 🗓️ Model lifecycle — a deprecation calendar with links to provider announcements.

Related MCP server: Tickerr - Live AI Tool Status & API Pricing

🔧 Built for measurement

  • Python standard-library core for edge probes; provider keys are needed for inference probes.

  • SQLite storage and regional aggregation, with remote nodes shipping measurements to a central node.

  • Static publishing for the website and JSON datasets.

  • CI checks for tests and linting; published snapshots include their measurement date.

Reading the results: edge TTFB measures network/API responsiveness, not model generation speed. Compare it separately from inference TTFT; results depend on region, probe type and time window.

🚀 Quick start

Use the public data without installing anything:

curl https://llmlatency.dev/api/rankings.json

Or run edge probes locally with Python 3.12+:

git clone https://github.com/mazamaka/llm-latency-tracker.git
cd llm-latency-tracker
REGION=local python3 run.py
python3 aggregate.py --region local

For a local stdio MCP server over the published API (stdlib only, no keys):

python3 mcp_server.py

Tools: get_ai_api_latency (provider leaderboard per region), get_provider_latency (one provider across regions), get_model_deprecations (model shutdown calendar), get_coding_agent_health (Claude Code / Codex speed and accuracy drift). All read-only.

Inference probes, API access & site generation → · Deployment →

🧪 Contributing

Add a provider, bring a new probe region, or improve the measurement code. See CONTRIBUTING.md for setup and checks.


Python · SQLite · JSON API · MCP · Cloudflare Pages

Code: MIT · Data: CC-BY-4.0, with attribution to llmlatency.dev.

Daily snapshot — 2026-11-21

Measured latency across 46 AI inference providers in 4 regions. Method: distributed edge (DNS→TCP→TLS→TTFB) + inference (TTFT) probes, last 24h. License: CC-BY-4.0.

Region

Fastest provider (p50)

p50

p95

Uptime

Asia (Tokyo)

fireworks

19 ms

73 ms

100%

Europe (Germany)

geometric

10 ms

13 ms

100%

South America (Santiago)

openrouter

55 ms

85 ms

100%

US (Central)

fireworks

29 ms

68 ms

100%

Snapshot generated 2026-11-21T07:46:59Z — this table is regenerated daily.

Available Tools

4 tools
get_ai_api_latencyAI API latency leaderboard by regionA
Read-onlyIdempotent
Inspect

Rank hosted AI inference API providers (46 tracked: OpenAI, Anthropic, Google, Groq, Fireworks, OpenRouter and others) by measured network latency inside a probe region over the last 24 hours. Use it to answer 'which provider responds fastest from Europe/US/Tokyo/Sao Paulo' or to pick an endpoint close to your users. To follow one named provider across regions use get_provider_latency instead. Values are edge TTFB (DNS+TCP+TLS+first byte from probes every 5 minutes, direct connections, no gateway), not tokens per second or model quality. Returns the region label, dataset timestamp and the ranked rows; omit region to get all four regions at once (larger response).

Inputs: limit, region. Returns: source, window, regions, generated

get_coding_agent_healthClaude Code / Codex speed and accuracy driftA
Read-onlyIdempotent
Inspect

Report whether the Claude Code or Codex CLI coding agents (on consumer subscriptions) got slower, think less or answer worse than their own recent baseline. Gives output speed (tokens/s and ms per token), thinking tokens, time to first token and error rate for the last 24 h versus the previous 7 days with Mann-Whitney significance flags, plus accuracy on a fixed hard question panel. Use it when someone asks 'is Claude Code slow / nerfed today'. Measured hourly from one EU server, one account per vendor, so it reflects that vantage point; for raw API latency use get_ai_api_latency.

Inputs: agent. Returns: rules, agents, vantage, generated

get_model_deprecationsAI model deprecation and shutdown calendarA
Read-onlyIdempotent
Inspect

List announced shutdowns of AI models (OpenAI, Anthropic, Mistral, Azure OpenAI and others) with announcement date, shutdown date, recommended replacement and the provider page each entry was verified against, plus how many days of notice each provider typically gives (median/min/max). Use it before pinning a model version, to plan migrations, or to check whether a model ID still works. Returns upcoming shutdowns by default; set include_past to also get already-retired models. This is calendar data, not latency; for speed use get_ai_api_latency.

Inputs: provider, include_past. Returns: past, source, upcoming, generated, +1 more

get_provider_latencyOne AI provider's latency in every regionA
Read-onlyIdempotent
Inspect

Show how a single AI API provider performs from every probe region (Germany, USA, Tokyo, Sao Paulo): its rank in that region, median and p95 time to first byte, uptime and sample count over the last 24 hours. Use it for questions like 'is the Anthropic API slower from Asia than from Europe' or 'was OpenAI reachable today'. For the full leaderboard of a region use get_ai_api_latency. An unknown provider id returns an error that lists the valid ids.

Inputs: provider (required). Returns: source, regions, provider, generated

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.5/5.0

    Scored across 4 tools

    Disambiguation5/5

    The two latency tools have overlapping subject matter, but each description explicitly points to the other with a clear boundary rule (leaderboard vs. single-provider across regions). The deprecations and coding-agent-health tools are in entirely different domains, and each description reiterates its scope ('calendar data, not latency', 'for raw API latency use...'). An agent can pick correctly without guessing.

    Naming Consistency5/5

    All four tools use the same get_<subject> snake_case pattern with no deviation in verb or casing. The names are predictable and readability is high. Minor quibble: get_ai_api_latency actually returns a ranked leaderboard rather than a single latency value, but the convention itself is uniform.

    Tool Count4/5

    Four tools is compact and none feel redundant, which fits a narrowly scoped observability server. It is on the lean side given it spans three distinct areas (API latency, model deprecations, coding-agent health), leaving little room for supporting lookups.

    Completeness4/5

    The surface covers the core questions: regional ranking, per-provider detail, deprecation planning, and agent regression checks. Gaps are minor and workaroundable (no provider-list endpoint, though errors enumerate valid ids; no historical time series beyond the fixed 24h window).

    Maintenance

    ActivityActive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      D
      maintenance
      Real-time status monitoring, uptime tracking, incident history, and API pricing for 42+ AI tools including ChatGPT, Claude, Gemini, Cursor, GitHub Copilot, Perplexity, DeepSeek, and Groq. No API key required. Data updated every 5 minutes from independent monitoring infrastructure.
      7
      58 npm
      1
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      Live, reproducible crypto infrastructure benchmarks for AI agents: RPC latency, bridge fees, L1 finality, gas oracle accuracy, stablecoin pegs. Three tools (list_benchmarks, get_benchmark,query_prom), no API key, CC-BY-4.0 data.
      11
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Enables AI coding assistants and agent workflows to query live measured free-LLM performance and availability, including throughput, uptime, tool-calling success, and model diagnostics.
      4
      2
      MIT