# tensorzero/tensorzero

TensorZero is an open-source LLMOps platform that unifies an LLM gateway, observability, evaluation, optimization, and experimentation.

Repository: https://github.com/tensorzero/tensorzero
Canonical: https://ross.abutalabs.com/products/tensorzero
Homepage: https://tensorzero.com
Language: Rust
License: Apache-2.0
License Family: permissive
Topics: ai, artificial-intelligence, deep-learning, gpt, llm, llmops, llms, machine-learning, rust, ml, mlops, anthropic, llama, openai, generative-ai, ai-engineering, python, ml-engineering, large-language-models, genai
Archived: true
Last push: 2026-06-11T01:48:44+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 86, release rhythm 75, longevity 55
- inputs: {"age_days": 778, "days_push": 84, "days_rel": 90, "gap_med": 4.0, "n_releases_24m": 121}
- flags: archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 11719, forks 962 (observed 2026-08-28T04:10:49.571479+00:00)

## What it is
TensorZero is an open-source LLMOps platform that unifies an LLM gateway, observability, evaluation, optimization, and experimentation behind a single high-performance API. It is written in Rust, works with the OpenAI SDK and every major LLM provider, and can be adopted incrementally.

## Use cases
- route requests to multiple llm providers through one gateway
- log and store llm inferences and feedback for observability
- evaluate llm outputs with heuristics and llm judges
- optimize prompts and models from production metrics
- run a/b tests and experiments on llm applications
- add fallbacks and retries to llm api calls
- monitor llm application performance in production

## When to choose
- you need a unified, low-latency API across many LLM providers
- you want production observability and feedback loops for LLM apps
- you need systematic evaluation and A/B testing of prompts and models
- you want to optimize LLM systems using real production data

## When to avoid
- you only need a simple single-provider SDK without ops features
- you want a fully managed SaaS with no self-hosting responsibility
- your stack cannot run Rust-based services or a gateway component

## Facets
- artifact type: service
- maturity: active
- function: llm-inference, monitoring, api-gateway, benchmarking, prompt-engineering, rag, agent-framework
- domain: large-language-models, artificial-intelligence, machine-learning, developer-tools
- platform: rust, python, self-hosted, cloud, cross-platform
- tags: llmops, llm-gateway, observability, a-b-testing, evaluation, model-optimization, openai-compatible, opentelemetry, mlops, docker

## Member repositories
- tensorzero/tensorzero (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:49.571479+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:15:22.870865+00:00, confidence not recorded.
  - readme: https://github.com/tensorzero/tensorzero (fetched 2026-08-28T04:10:49.571479+00:00, sha 10bf8ee1b985)
  - homepage: https://tensorzero.com (fetched 2026-08-29T08:13:36.575807+00:00, sha f85ff86e48c2)
  - registry_crates: https://crates.io/api/v1/crates/tensorzero (fetched 2026-08-29T08:13:36.578245+00:00, sha 51d5fafcc9d4)
- Data as of 2026-08-30T08:39:29.467469+00:00.