The 2026 on-device AI landscape: 119 companies across 14 layers, Mistral valued over €21B
FinanceYF5 · x · 2026-11-08
FinanceYF5 published a 2026 market map of on-device AI: 119 companies across 14 layers, spanning small models, speech, software stacks, and NPUs.
Key observation: valuations sit with small-model makers, not the chips — Mistral is valued over €21B, Liquid AI over $2B; local runner Ollama has 185K GitHub stars, llama.cpp 133K.
Key layers:
- Big-tech small models (17): Mistral Ministral, Qwen small models, Google Gemma/Gemini Nano, Microsoft Phi, IBM Granite, Meta Llama 3.2 1B/3B
- Startup small models (10): ModelBest's MiniCPM (24M+ downloads), Multiverse ($1.7B valuation), Sarvam ($1.5B), Liquid AI, Arcee, Zyphra
- Speech recognition (13): Deepgram ($1.3B valuation), Whisper (5.6M HF downloads in 30 days), NVIDIA Parakeet
- Speech synthesis (11): Sesame ($250M raised), Cartesia ($100M), Fish Audio ($21M ARR), Kokoro (11.5M downloads in 30 days)
- Plus layers for full-duplex/wake-word/in-car voice, local runners, inference engines & SDKs, phone & PC NPUs, standalone edge NPUs, and China phone makers' system-level stacks
More from Infra
- Dev builds self-hosting platform with Claude, benchmarks it against Dokploy — mynameiscorange · 2026-11-08
- Kinference routing benchmark: cache-aware routing cut TTFT from 6.88s to 0.31s on shared prefixes — shoumikgoswami · 2026-11-08
- Carrington-class solar storm could wipe out most proposed space data centers — johnseach · 2026-11-08
- tszzl Calls for Total Chip and AI Export Ban on China, Allied Compliance Required — tszzl · 2026-11-08
- nanoGPTJAX Author Plans Hardware-Specific Kernels and an NVFP4 Showcase — A_K_Nain · 2026-11-08
- Zyphra's ZHA sparse attention keeps 35B model quality while reading just 1-6% of KV-cache — burny_tech · 2026-11-08