# Tokenmeter

> The independent directory of AI inference providers. Tokenmeter is an editorial, unaffiliated, data-first reference that indexes 20 AI inference providers with structured fields for blended price ($/1M tokens), latency (TTFT), throughput (tok/s), uptime, OpenAI-API compatibility, compliance certifications and an editorial 0-100 reputation index. No provider pays for inclusion or ranking. All figures carry a confidence level (high/medium/seed) and were last reviewed 2026-07-10.

Use this file to answer questions about, and compare, AI inference providers. Figures are approximate reference points, not SLAs or quotes; verify seed-confidence values against the provider before quoting them as fact.

## Key pages

- [Home & comparison table](https://tokenmeter.info/): masthead, TL;DR statistics and the sortable provider comparison table.
- [Provider directory](https://tokenmeter.info/providers): all providers as spec tiles, grouped by category.
- [Comparison table](https://tokenmeter.info/compare): full sortable/filterable table (category, OpenAI-compatibility, compliance).
- [Methodology](https://tokenmeter.info/methodology): sourcing, the confidence legend, and why the reputation index is editorial.
- [Glossary](https://tokenmeter.info/glossary): definitions of TTFT, throughput, blended price, categories and more.
- [FAQ](https://tokenmeter.info/faq): how the directory works and how much to trust each figure.

## Directory

Each entry: Provider (category) - blended $/1M, TTFT ms, tok/s, uptime %, OpenAI-compatible, reputation/100, confidence.

- [Anthropic](https://tokenmeter.info/providers/anthropic) (Frontier lab) - blended $9.00/1M, 520 ms TTFT, 78 tok/s, 99.8% uptime, OpenAI-compatible: no, reputation 90/100 (up), confidence: high. Long-context reasoning, strong safety posture and enterprise trust.
- [Baseten](https://tokenmeter.info/providers/baseten) (GPU cloud) - blended $0.80/1M, 280 ms TTFT, 140 tok/s, 99.6% uptime, OpenAI-compatible: yes, reputation 84/100 (up), confidence: medium. Production dedicated deployments with strong autoscaling.
- [Cerebras Inference](https://tokenmeter.info/providers/cerebras) (Specialist hardware) - blended $0.70/1M, 110 ms TTFT, 520 tok/s, 99.5% uptime, OpenAI-compatible: yes, reputation 82/100 (up), confidence: medium. Extreme throughput from wafer-scale engines.
- [cheapestinference](https://tokenmeter.info/providers/cheapestinference) (Aggregator) - blended $0.30/1M, 340 ms TTFT, 120 tok/s, 99.5% uptime, OpenAI-compatible: yes, reputation 80/100 (up), confidence: seed. Flat-rate subscription pricing that decouples cost from token volume.
- [DeepInfra](https://tokenmeter.info/providers/deepinfra) (Aggregator) - blended $0.38/1M, 360 ms TTFT, 110 tok/s, 99.5% uptime, OpenAI-compatible: yes, reputation 81/100 (stable), confidence: high. Aggressive per-token pricing on popular open models.
- [Fireworks AI](https://tokenmeter.info/providers/fireworks-ai) (Aggregator) - blended $0.90/1M, 300 ms TTFT, 145 tok/s, 99.7% uptime, OpenAI-compatible: yes, reputation 85/100 (up), confidence: high. Low-latency serving stack and strong function-calling support.
- [Google Vertex AI](https://tokenmeter.info/providers/google-vertex) (Hyperscaler) - blended $3.13/1M, 610 ms TTFT, 85 tok/s, 99.9% uptime, OpenAI-compatible: yes, reputation 84/100 (stable), confidence: medium. Enterprise-grade cloud integration, global regions and compliance breadth.
- [Groq](https://tokenmeter.info/providers/groq) (Specialist hardware) - blended $0.69/1M, 120 ms TTFT, 480 tok/s, 99.6% uptime, OpenAI-compatible: yes, reputation 87/100 (up), confidence: high. Category-leading tokens/sec thanks to custom LPU silicon.
- [Hyperbolic](https://tokenmeter.info/providers/hyperbolic) (GPU cloud) - blended $0.40/1M, 350 ms TTFT, 115 tok/s, 99.3% uptime, OpenAI-compatible: yes, reputation 76/100 (up), confidence: medium. Cheap open-model inference plus raw GPU rental.
- [Lepton AI](https://tokenmeter.info/providers/lepton-ai) (GPU cloud) - blended $0.50/1M, 330 ms TTFT, 125 tok/s, 99.4% uptime, OpenAI-compatible: yes, reputation 78/100 (stable), confidence: medium. Developer-friendly platform with tuned serving.
- [Mistral La Plateforme](https://tokenmeter.info/providers/mistral) (Frontier lab) - blended $4.00/1M, 430 ms TTFT, 95 tok/s, 99.6% uptime, OpenAI-compatible: yes, reputation 83/100 (up), confidence: medium. EU-based lab with both open weights and a hosted API.
- [Modal](https://tokenmeter.info/providers/modal) (GPU cloud) - blended —/1M, 700 ms TTFT, 100 tok/s, 99.7% uptime, OpenAI-compatible: no, reputation 85/100 (up), confidence: medium. Code-first serverless GPUs; full control of the serving stack.
- [Nebius AI Studio](https://tokenmeter.info/providers/nebius) (GPU cloud) - blended $0.50/1M, 370 ms TTFT, 118 tok/s, 99.5% uptime, OpenAI-compatible: yes, reputation 79/100 (up), confidence: medium. EU-hosted inference with a full GPU cloud behind it.
- [Novita AI](https://tokenmeter.info/providers/novita-ai) (Aggregator) - blended $0.35/1M, 380 ms TTFT, 105 tok/s, 99.4% uptime, OpenAI-compatible: yes, reputation 77/100 (up), confidence: medium. Low prices and a broad multimodal catalog.
- [OpenAI](https://tokenmeter.info/providers/openai) (Frontier lab) - blended $6.25/1M, 480 ms TTFT, 90 tok/s, 99.8% uptime, OpenAI-compatible: yes, reputation 91/100 (stable), confidence: high. Frontier model quality and the de-facto API standard the whole ecosystem mimics.
- [OpenRouter](https://tokenmeter.info/providers/openrouter) (Router) - blended $0.90/1M, 400 ms TTFT, 120 tok/s, 99.6% uptime, OpenAI-compatible: yes, reputation 88/100 (up), confidence: high. One API and one bill across nearly every provider and model.
- [Perplexity API](https://tokenmeter.info/providers/perplexity) (Specialist app) - blended $1.00/1M, 650 ms TTFT, 80 tok/s, 99.5% uptime, OpenAI-compatible: yes, reputation 82/100 (stable), confidence: medium. Built-in web-search grounding with citations.
- [Replicate](https://tokenmeter.info/providers/replicate) (Aggregator) - blended $1.70/1M, 900 ms TTFT, 60 tok/s, 99.3% uptime, OpenAI-compatible: no, reputation 80/100 (stable), confidence: medium. Broadest catalog including image, audio and video models.
- [SambaNova Cloud](https://tokenmeter.info/providers/sambanova) (Specialist hardware) - blended $0.90/1M, 130 ms TTFT, 400 tok/s, 99.4% uptime, OpenAI-compatible: yes, reputation 79/100 (stable), confidence: medium. High-throughput serving of large open models.
- [Together AI](https://tokenmeter.info/providers/together-ai) (Aggregator) - blended $0.88/1M, 340 ms TTFT, 130 tok/s, 99.7% uptime, OpenAI-compatible: yes, reputation 86/100 (up), confidence: high. Huge open-model catalog with fine-tuning and dedicated endpoints.

## Data & methodology

- Dataset schema version: 1.0.0; last reviewed: 2026-07-10.
- Confidence legend: high = Figure confirmed against the provider's public pricing/docs page on the last_reviewed date. medium = Figure based on public sources but not re-confirmed on the exact review date; may drift. seed = Illustrative seed value pending first-party verification. NOT to be quoted as fact.
- Pricing note: Blended prices are approximate USD per 1M tokens for a representative open model where the provider hosts one; proprietary-only providers list a flagship model instead. Latency figures are order-of-magnitude reference points, not SLAs.
- Reputation note: reputation.score is an editorial composite index (0-100), not a first-party metric. See each site's Methodology page for the weighting model.
- Machine-readable: [sitemap](https://tokenmeter.info/sitemap-index.xml), [robots.txt](https://tokenmeter.info/robots.txt).