Builds

Dated session write-ups from real desk runs — one-line result on every card.

Next: open a card below, or take the numbers desk at toklanes.com/reviews. Method notes: Method notes.

All sessions

Gemma12-5060 GrokBuild matrix

2026-09-30 · speed 20260930-150835-1 · quality 20260930-121551 · RTX 5060 Ti

Result: speed eng 88.18 tok/s · wall 39.59 tok/s · ok (150835); quality still 121551 — bugfix pass 1/1; tooljson pass 4/4; scriptreview partial 3/6; refactor ok.

B70

B70 vs 5080 — text, vision, gen blocked

2026-09-21 · rumble · B70 SYCL · 5080 CUDA

Result: Phase B B70 text 4×3 pass; Phase C mixed; img describe B70 FAIL / 5080 PASS; image/video gen BLOCKED.

Lanes

Local LLM lanes: speed, capacity, and the hybrid that isn’t

2026-09-09 · 5080 · B70 · Spark · Halo · Mac

Result: Speed lane = B70-class ≤32GB; capacity = Spark/Halo/Mac ~128GB — not one hybrid SKU.

Compare

Quality tied — speed and VRAM decide

2026-09-08 · RTX 5080

Result: Gemma 12B ~70 tok/s vs Qwen 27B IQ4 ~28 tok/s — same 4/4 quality score.

MoE

FreeToken on WSL2 — mid MoE yes, Flash-Next no

2026-09-07 → 2026-09-08 · MoE · 5080

Result: 35B-A3B smoked (wall ~60–90); Flash-Next did not fit. Negative results count.

Quant

Same 27B, three quants — only one clears the bar

2026-09-08 · Unsloth · Ollama · 5080

Result: IQ4_XS full-GPU ~43 tok/s; library Q4 ~18; 46-layer spill ~7.

Editor

When the model “thinks,” the editor sees blank

2026-09-02 → 2026-09-08 · VS Code

Result: Answers landed in reasoning fields; think-off aliases fixed the UX.

Research

When the research swarm reports back

2026-09-09 · last30days

Result: One word (Intel), many public sources, one usable hardware map.

Thumbnails are placeholders (not run yet / no fake video IDs). Full write-ups are the source of truth.