DiggTop · the local-AI hardware leaderboard
The local AI hardware leaderboard.
Buy with data, not hype — community benchmark runs, VRAM-first guides and weekly street prices. Three steps, real numbers, before you buy.
Live signals · Oct 19, 2026
491benchmark runs
359verified
77GPUs tracked
Oct 17, 2026price snapshot
Verified street prices
Oct 17, 2026 snapshot
| GPU | Market | Price | $/GB |
|---|---|---|---|
| Intel Arc Pro B70 | Newegg | $1,300 | $41 |
| RTX 3090 | Us Street Low | $2,050 | $85 |
Benchmark HQ
BenchmarksVerified leaderboards and attributed runs for the GPUs and models people actually run at home.
Compare GPUsSide-by-side verified matrix for any GPU pair — zero JavaScript, just pick two.RTX 3090 vs 5090 →Street pricesWeekly snapshots of new, used and rental GPU prices with per-vendor sourcing.As of Oct 17, 2026 →- RTX PRO 6000 Blackwell1271.3
- Radeon AI Pro R9700969.4
- RTX 3090709.5
Run Session · submit in 3 moves
Run it locally. Show the world.
No downloads, no scripts. Copy one prompt to your local AI — it measures and posts three steps live; you confirm, then share a run card stamped DIGGTOP.
- Copy the prompt
- Send to your local AI
- Confirm & share
Latest analysis
- CMP 170HX, B70, V100: Cheap Cards Punching UpSame model, same W4A16 quant, same vLLM engine: a $1,050 CMP 170HX runs 255.9 tok/s and a $9,000 RTX 5090 gets 137.9.Sep 28, 2026
- Qwen3.8-27B on 22 GPUs: 44 Sourced RunsOne model, 22 GPUs, 44 sourced runs, 16.8 to 435 tok/s. The card matters less than the quant, the engine and the context you run it at.Sep 28, 2026
- One 180B MoE, Five Formats: 41 Sourced RunsQwen3.8-Flash-Next's split entities merged into one page: 41 sourced rows across three checkpoints, 20.3 to 474 tok/s, and what fits a 24GB card.Sep 28, 2026
Latest guides
- Local LLM GPU Buying Guide (2026)What to buy to run local LLMs in September 2026, chosen by VRAM tier with current US prices, used-market alternatives and honest tradeoffs.Sep 4, 2026
- How to read the DiggTop boardWhat the checkmark, the self tag and the ellipsis mean on our benchmark board, and why some imported numbers stay partial.Sep 10, 2026
- What $1,000 buys for local LLMs (Sep 2026)What $500-$1,000 buys in VRAM for local LLMs, from the 2026-09-28 price snapshot: 12GB, 16GB, 24GB and now 32GB.Sep 10, 2026
- The 48GB VRAM Local LLM BuildHow to get 48GB of VRAM for 70B-class local LLMs: dual RTX 3090 vs used A6000 48GB — power, cooling, PCIe and real community tok/s.Sep 5, 2026
- Best GPU for Qwen3: VRAM Tiers per ModelWhich GPU runs which Qwen3 model — 0.5B to 180B mapped to VRAM tiers, with real community tok/s from 143 sourced DiggTop benchmark runs.Sep 5, 2026
Fresh builds
- Dual RTX 3090 70B Build (48GB, 2026)Two used 24GB cards give 48GB for 70B-class local inference. What it costs at the 2026-09-28 snapshot, and sourced tok/s.Sep 6, 2026
- RTX 5090 32GB Flagship Build (2026)The 2026 single-card speed ceiling: an RTX 5090 32GB build against the 2026-09-28 price snapshot, with sourced-only board numbers.Sep 6, 2026
- $1,500 Sweet-Spot Build (Used RTX 3090)The best tokens-per-dollar in 2026: one used RTX 3090 24GB runs 32B models interactive and 70B at patient speeds. Prices as of 2026-09-28.Sep 4, 2026