# Provider: open-inference

- **TrackLLM** — machine-readable twin of [providers/open-inference.html](https://www.trackllm.net/providers/open-inference.html). Every page on the site has a `.md` sibling at the same path; links below point at those.
- **Atom feed** of this page's detected changes: [feeds/providers/open-inference.xml](https://www.trackllm.net/feeds/providers/open-inference.xml)
- **JSON** this page is rendered from: [data/providers/open-inference.json](https://www.trackllm.net/data/providers/open-inference.json) (provider JSON)
- **Raw data** (every collected response): https://github.com/timothee-chauvin/trackllm_data · Built 2026-10-15 18:26 UTC

[Home](https://www.trackllm.net/index.md) / open-inference

6 endpoints across 5 models and 3 serving variants, monitored since 2026-07-17. 2 changes detected across 2 endpoints.

## Drift rate

Detected changes per endpoint-year of monitoring, with a 95% Poisson interval. A rate is withheld below 0.5 endpoint-years of exposure.

| Method | Endpoints | Endpoint-years | Changes | Rate (/endpoint-year) | 95% interval |
| --- | --- | --- | --- | --- | --- |
| LT | 2 | 0.1 | 0 | — | — |
| B3IT | 6 | 0.55 | 2 | 3.63 | 0.00–8.67 |

## Drift by model

Every endpoint open-inference serves. Each endpoint's full drift series and detected changes are in the JSON above.

## Monitoring & changes over time

Endpoints under monitoring each month (the exposure the rate divides by) are in the JSON above under `months` and `variants[].monitoring`.

## Serving variants

A provider's quantizations are separate serving stacks and drift separately.

| Variant | Endpoints | LT rate | LT 95% interval | LT changes | B3IT endpoints | B3IT changes |
| --- | --- | --- | --- | --- | --- | --- |
| fp4 | 4 | — | — | 0 | 4 | 1 |
| bf16 | 1 | — | — | 0 | 1 | 0 |
| fp8 | 1 | — | — | 0 | 1 | 1 |

## Endpoints

| Model | Variant | Org | Status | Changes | Methods | Last change |
| --- | --- | --- | --- | --- | --- | --- |
| [deepseek-v4-flash](https://www.trackllm.net/endpoints/deepseek2fdeepseek-v4-flash23open-inference2ffp8.md) | [open-inference/fp8](https://www.trackllm.net/providers/open-inference.md) | [deepseek](https://www.trackllm.net/orgs/deepseek.md) | retired | 1 | `LT` `B3IT` | 2026-09-22 |
| [deepseek-v4.1-flash](https://www.trackllm.net/endpoints/deepseek2fdeepseek-v4.1-flash23open-inference2ffp4.md) | [open-inference/fp4](https://www.trackllm.net/providers/open-inference.md) | [deepseek](https://www.trackllm.net/orgs/deepseek.md) | tracked | 1 | `B3IT` | 2026-09-26 |
| [deepseek-v4-flash](https://www.trackllm.net/endpoints/deepseek2fdeepseek-v4-flash23open-inference2ffp4.md) | [open-inference/fp4](https://www.trackllm.net/providers/open-inference.md) | [deepseek](https://www.trackllm.net/orgs/deepseek.md) | tracked | 0 | `B3IT` | — |
| [deepseek-v4-flash-0731](https://www.trackllm.net/endpoints/deepseek2fdeepseek-v4-flash-073123open-inference2ffp4.md) | [open-inference/fp4](https://www.trackllm.net/providers/open-inference.md) | [deepseek](https://www.trackllm.net/orgs/deepseek.md) | tracked | 0 | `B3IT` | — |
| [gemma-4-31b-it](https://www.trackllm.net/endpoints/google2fgemma-4-31b-it23open-inference2fbf16.md) | [open-inference/bf16](https://www.trackllm.net/providers/open-inference.md) | [google](https://www.trackllm.net/orgs/google.md) | retired | 0 | `LT` `B3IT` | — |
| [glm-5.3-flash](https://www.trackllm.net/endpoints/z-ai2fglm-5.3-flash23open-inference2ffp4.md) | [open-inference/fp4](https://www.trackllm.net/providers/open-inference.md) | [z-ai](https://www.trackllm.net/orgs/z-ai.md) | tracked | 0 | `B3IT` | — |
| [deepseek-v4-flash-0731](https://www.trackllm.net/endpoints/deepseek2fdeepseek-v4-flash-073123open-inference2ffp8.md) | [open-inference/fp8](https://www.trackllm.net/providers/open-inference.md) | [deepseek](https://www.trackllm.net/orgs/deepseek.md) | retired | 0 | — | — |
| [gpt-oss-120b](https://www.trackllm.net/endpoints/openai2fgpt-oss-120b23open-inference2fint8.md) | [open-inference/int8](https://www.trackllm.net/providers/open-inference.md) | [openai](https://www.trackllm.net/orgs/openai.md) | retired, errors out | 0 | — | — |

## Detected changes

| Date | Endpoint | Method | Magnitude | Description |
| --- | --- | --- | --- | --- |
| 2026-09-26 | [deepseek-v4.1-flash @ open-inference/fp4](https://www.trackllm.net/endpoints/deepseek2fdeepseek-v4.1-flash23open-inference2ffp4.md) | B3IT | TV 0.39 | Probe-prompt outputs shifted (total variation 0.39) |
| 2026-09-22 | [deepseek-v4-flash @ open-inference/fp8](https://www.trackllm.net/endpoints/deepseek2fdeepseek-v4-flash23open-inference2ffp8.md) | B3IT | TV 0.52 | Probe-prompt outputs shifted (total variation 0.52) |


---
Data collected via the OpenRouter API. [Home](https://www.trackllm.net/index.md) · [Changes](https://www.trackllm.net/changes.md) · [Providers](https://www.trackllm.net/providers.md) · [Endpoints](https://www.trackllm.net/endpoints.md) · [Methodology](https://www.trackllm.net/methodology.md) · [About](https://www.trackllm.net/about.md) · [GitHub](https://www.trackllm.net/github.md)
