Model supportv2.0Aug 3, 2026
Qwen 3.5 and Gemma 4 support
Photon 2.0 adds production support for Qwen 3.5 (0.8B-9B) and Gemma 4 E2B/E4B.
- Models
- Qwen 3.5 0.8B · Qwen 3.5 2B · Qwen 3.5 4B · Qwen 3.5 9B · Gemma 4 E2B · Gemma 4 E4B
- Hardware
- H100
Hardware supportv2.0Aug 3, 2026
H100 80GB support
Photon 2.0 is tested on H100 SXM 80GB across all launch models.
- Models
- All models
- Hardware
- H100
- Benchmarks
- Photon was ahead in all 52 matched throughput comparisons: 28 against vLLM and 24 against SGLang, from 1.01x to 2.33x on ChartQA at concurrency 1-8.
Performancev2.0Aug 3, 2026
Lower cold-start time
Matched cold-start tests reached first response faster on Photon than vLLM and SGLang for the launch models.
- Models
- All models
- Hardware
- H100
- Benchmarks
- Photon had up to 66% lower total cold-start time than vLLM across seven models and up to 55% lower time than SGLang across six supported models.
Capabilityv2.0Aug 3, 2026
Moondream 3 segmentation
Moondream 3 on Photon serves segmentation masks alongside query, caption, detect, and point.
- Models
- Moondream 3
- Hardware
- H100
Runtimev2.0Aug 3, 2026
Compiled megakernel execution
Request handling, tokenization, and image decode stay on the CPU. Model execution runs inside one compiler-generated GPU program.
- Models
- All models
- Hardware
- All hardware
Compatibilityv2.0Aug 3, 2026
md.photon() requires moondream SDK 0.11+
md.photon() needs moondream 0.11 or later. Earlier SDKs stay on Moondream-only local inference.
- Models
- All models
- Hardware
- All hardware
Deprecationv2.0Aug 3, 2026
Photon 1.x edge support ended
Moondream 2 on A10 and Jetson AGX Orin is no longer supported for new deployments.
- Models
- Moondream 2
- Hardware
- A10 · Jetson AGX Orin
Known limitations- Photon 1.x configurations receive critical fixes only.
