Running a 35B LLM at 128K Context, Full Speed, on €870 of Used Hardware
Every number below is measured on the box, not estimated.Two phases, two bottlenecks. Prefill (processing your prompt) is compute-bound — it crunches the whole prompt in parallel and hits hundreds-to-thousands of tokens/s. Decode (writing the answer, one token at a time) is memory-bandwidth-bound — every token re-reads the active weights from VRAM.
- ▪Every number below is measured on the box, not estimated.Two phases, two bottlenecks.
- ▪Prefill (processing your prompt) is compute-bound — it crunches the whole prompt in parallel and hits hundreds-to-thousands of tokens/s.
- ▪Decode (writing the answer, one token at a time) is memory-bandwidth-bound — every token re-reads the active weights from VRAM.
Hacker News (AI / LLM) files mainly under ai. We currently carry 3,208 of its stories.
Story provenance
Source · retrieval · rights · ranking — open for full record
inspect →
Story provenance
Attribution is not the same as permission. This drawer separates discovery metadata, excerpts, WeSearch-generated summaries, reuse status, and whether the publisher receives the visit. Nothing here claims a legal grant the publisher has not made.
Record
| Original publisher | Medium |
| Canonical URL | https://medium.com/ai-advances/running-a-35b-llm-at-128k-context-full-speed-on-870-of-used-hardware-no-cloud-required-c4f7629810b8 |
| Publication time | Sun, 02 Aug 2026 08:27:38 +0000 |
| Retrieval time | 2026-08-02T08:30:40.489Z |
| Last seen | 2026-08-02T08:30:40.489Z |
| Headline source | Publisher (no WeSearch rewrite) |
| Excerpt source | publisher body |
| Excerpt method | First ~120 words (~800 chars) of extracted publisher body, fair-use limited. |
| Summary | WeSearch · cerebras-chat (WeSearch summarizer) |
| Summary source text | contentText |
| Citation coverage | Summary is a WeSearch-generated derivative; primary citation is the original publisher URL. |
| Cluster | UGcW03TvTiTP · 1 stories |
| Cluster logic | Grouped by semantic title/content similarity across sources within a rolling window. Same-publisher template collisions are excluded from coverage comparison. |
| Ranking reason | Story pages are not engagement-ranked. Hub feeds use recency, with optional source-diversified chronological ordering (cap consecutive stories per source). No personalized ranking. |
| Publisher visit | Yes — open original |
| Substitutes article? | No — link-out required for full text |
Rights status (four layers)
WeSearch handling by dimension
| Indexing | May the item be indexed (stored, ranked, made findable)? | Allowed |
| Snippet | May a short excerpt of the publisher's text be shown? | Allowed |
| AI summary | May WeSearch generate its own short summary of the article? | Limited |
| Retrieval / RAG | May the content be exposed for third-party retrieval-augmented generation? | Not asserted |
| Model training | May the content be used to train AI models? | Not asserted |
| Commercial reuse | May the content be reused commercially? | Not permitted |
Basis: Derived from the published RSS/Atom feed. Contact: [email protected]. Reviewed: 2026-07-24.
Opening excerpt (first ~120 words) tap to expand
Local LlmArtificial IntelligenceAIOllamaLlama CppRunning a 35B LLM at 128K Context, Full Speed, on €870 of Used Hardware — No Cloud RequiredGian Luca Bailo14 min read·Jul 16, 2026--5ListenShareHow a €870 used-GPU box runs a 35B model at full speed all the way to 128K context — and what I learned measuring the upgrade, including the pricier card that made my smaller models slower, not faster.Press enter or click to view image in full sizeOriginal hardware, msi 4070 12G and asus 5060ti 16GTwo used-GPU builds, five frontier models, both fully benchmarked — and the answer to whether doubling the GPU budget doubles anything.I run a small inference box at home and I benchmarked it in two hardware configurations.The €600 build: two mismatched used GPUs bolted together, an RTX 4070 (12 GB, Ada,…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Medium.
Discussion
0 commentsMore from Medium
