128GB unified RAM laptop benchmarks: gemma4:e4b hits 122.8 TPS locally via Ollama
ollama · x · 2026-11-04
burkeholland benchmarked local models with Ollama on a new 128GB unified-memory laptop: gemma4:e4b reached 122.8 warm TPS with 8.65s cold TTFT, qwen3-coder:30b hit 90 TPS, gemma4:26b-a4b 86.9, gpt-oss:20b 60.7, and qwen3.8:latest 25.3. He notes it "feels really quick" for local inference on a laptop — useful reference data for on-device deployment.
More from Infra
- PE fund to pour $2B into a neocloud; SemiAnalysis vets its frontier-lab contract — AccBalanced · 2026-11-04
- DeepSeek serves only 17% of its own tokens as 19+ inference providers split the rest — AccBalanced · 2026-11-04
- A month at NetBird: agents devs ignore the networking layer at their peril — victorialslocum · 2026-11-04
- A buildout pause is actually good for hyperscalers: lower capex, higher returns, says analyst — pdamodaran · 2026-11-04
- Private fusion investment nears $10B as China catches up to US — StewartalsopIII · 2026-11-04
- 2.78T-param Kimi K3 runs on a single CPU with 8.24GB RAM in dependency-free C99 — FareedKhan-dev · 2026-11-04