About
Inference startup targeting agentic workloads. Serves open models like DeepSeek V4.1 Flash and GLM 5.3 Flash via endpoints that rank first on Artificial Analysis output-speed leaderboards. Their proprietary technique, DFlash 2, is a parallel block-diffusion speculative decoding method delivering over 20% more tokens per pass with roughly 1% added latency. Also ships Splash, an open-source local inference engine optimized for Apple silicon, running Qwen3 models at up to 3x Ollama's decode speed.
Activity
2026
2 stories so far in 2026
| Month | Stories |
|---|---|
| January | 0 |
| February | 0 |
| March | 0 |
| April | 0 |
| May | 0 |
| June | 0 |
| July | 0 |
| August | 1 |
| September | 0 |
| October | 1 |
| November | 0 |
Focus areas
2 stories
Inco AI Ships DFlash 2 Draft Model to Triple GLM-5.3-Flash Token Throughput
Inco AI's DFlash 2 drafter for GLM-5.3-Flash now ships under Apache 2.0, delivering near 3x speculative decoding speedups with lossless output.