One OpenAI-compatible endpoint. Every call routed to your fastest, cheapest provider. On your own keys.
Call Llama, DeepSeek, GPT-OSS, and Kimi from your existing OpenAI client. Scheduled benchmarks keep the scoreboard current, and every request routes to the winning provider.
The routing layer for the keys you already own
One OpenAI-compatible endpoint
Point your existing OpenAI client at Tesseract, pick a model, and go. It routes across every provider you have a key for — no per-provider client to maintain.
Bring your own provider keys
Add an API key for Together AI, Groq, Fireworks, Baseten, or OpenRouter. Keys are encrypted at rest, verified with a live test call, and only the last four are ever shown.
A live provider scoreboard
Tesseract benchmarks the same prompt across your connected providers and tracks which is fastest and cheapest per model — the data that powers routing.
Routing on real data
Every request is sent to the best provider under your objective — lowest latency, highest throughput, or lowest cost — using the scoreboard, not a guess.
Benchmarks that run themselves
Schedule recurring runs per model and the scoreboard stays current on its own — routing never rides on stale data.
Know when the leaderboard moves
Get notified when a provider degrades or a cheaper one appears for a model you care about.
Bring your keys. Point your client. Ship.
Connect your provider keys
Add an API key for each provider you already pay for. Tesseract encrypts it, verifies it with a live test call, and you only ever see the last four characters.
Point your client at Tesseract
The endpoint is OpenAI-compatible: swap the base URL, pick any catalog model, and your requests route through your own provider keys to the fastest, cheapest backend.
Watch the scoreboard, ship with confidence
Tesseract benchmarks every provider live and routes each request on the result. You pay your providers directly — Tesseract is the $39/mo routing layer on top.
client = OpenAI(
- base_url="https://api.openai.com/v1"
+ base_url="https://tesseract.click/api/v1"
api_key="your-tesseract-gateway-key",
)Same OpenAI SDK. Same model names. Inference runs on your provider keys; Tesseract routes each call to the fastest, cheapest one.
One catalog, every serving backend you bring.
Llama, DeepSeek, GPT-OSS, Kimi, and Qwen, all callable through one OpenAI-compatible endpoint on the keys you already own.
The catalogue refreshes nightly from each provider's live model list.
CATALOGUE LAST UPDATED 3h ago · SERVED BY 5/5 PROVIDERS
Gemma 3 4b It
Google · 4BLlama 3.2 3b Instruct
Llama · 3BLlama 3.2 1b Instruct
Llama · 1BLlama 3.1 8B
Llama · 8BMistral Small 24b Instruct 2501
Mistral · 24BGPT-OSS 20B
OpenAI · 20BQwen3.5 9B
Qwen · 9BQwen3 14b
Qwen · 14BDeepSeek V4 Flash
DeepSeek · 671BQwen3 Coder 30b A3b Instruct
Qwen · 30BQwen3 32b
Qwen · 32BGemma 4 26b A4b It
Google · 26BGpt Oss Safeguard 20b
OpenAI · 20BLlama 3.3 70B
Llama · 70BGemma 3 27b It
Google · 27BQwen 2.5 72B
Qwen · 72BLlama 3.1 70B
Llama · 70BGemma 4 31b It
Google · 31BQwen3.8 Flash
Qwen · —GPT-OSS 120B
OpenAI · 120BQwen3 30b A3b
Qwen · 30BGemma 2 27b It
Google · 27BQwen3 Vl 8b Instruct
Qwen · 8BQwen3.5 35b A3b
Qwen · 35BQwen2.5 Vl 72b Instruct
Qwen · 72BBeam 1.0 501b A23b
Beam · 501BDeepSeek V3
DeepSeek · 671BQwen3 Next 80b A3b Instruct
Qwen · 80BQwen3 Next 80b A3b Thinking
Qwen · 80BQwen3.7 Plus
Qwen · —Qwen3.8 27b
Qwen · 27BQwen3.6 Plus
Qwen · —Qwen3.6 27b
Qwen · 27BNemotron 3 Ultra 550b A55b
NVIDIA · 550BKimi K2.7 Code
Moonshot · —Kimi K2.6
Moonshot · —Qwen3.5 397b A17b
Qwen · 397BInkling
Inkling · —Qwen3.7 Max
Qwen · —Qwen3.8 2.4t A95b
Qwen · 4TKimi K3
Moonshot · 1TPER-MILLION-TOKEN RATES ARE EACH PROVIDER'S PUBLIC PRICING, SHOWN FOR TRANSPARENCY. YOU PAY YOUR PROVIDERS DIRECTLY; TESSERACT IS A FLAT $39/MO TOOL — NEVER PER-TOKEN METERING.
One tool. One flat monthly price.
Pro is $39/month for the routing and benchmarking tool. You bring your own provider keys and pay your providers directly for usage — Tesseract never marks up inference. Cancel anytime.
Free
Try the routing tool and live scoreboard. No credit card required.
- Connect up to 2 providers with your own keys
- OpenAI-compatible endpoint + routing on real benchmark data
- Live provider scoreboard per model
- Model catalogue with live per-provider pricing
- 1 Tesseract gateway key
- 5 benchmark runs per day
Pro
The routing and benchmarking tool, unlimited. You bring your own provider keys and pay your providers directly for usage.
- Connect unlimited providers with your own keys
- OpenAI-compatible endpoint + routing on real benchmark data
- Live provider scoreboard per model
- Custom routing rules per model
- Up to 5 Tesseract gateway keys
- Unlimited benchmark runs + scheduled benchmarks + alerts
Free tier included. Bring your own provider keys.
Stop guessing which provider to call.
Connect the keys you already have, point your OpenAI client at one endpoint, and let the live scoreboard route every request to the fastest, cheapest backend.
Start routing free in 2 minutes