AI Compute RadarGate: the smallest device with an EXCELLENT fit at 32K context. Seat: the measured GGUF file.
Pick of the week — Week 42, 2026 · selected by rule, written by people
Eight billion parameters that fit an eight-gigabyte card
LFM2.5 8B-A1B by Liquid AI takes the sixth pick: Heat Score 28 to 56 in seven days, the largest gain among models that run comfortably on a 24 GB card and are not on cooldown. Its Q4_K_M file measures 4.8 GB; at 8K context it needs 5.9 GB, which our engine rates GOOD on every 8 GB card we track.
Measured signals as of 2026-10-16 18:15 UTC · Hugging Face + Vast.ai · method on /methodology
Where it runs
Fit engine · measured GGUF + computed KV cache · usable memory after marginsOr rent an RTX 4090 from $1.07/h →Check it against your own machine →
What it is for
The rule is unchanged: the largest seven-day Heat Score gain among models that fit a 24 GB card with margin and were not picked in the last four weeks. Between the Friday runs of 9 and 16 October, both at 18:15 UTC, LFM2.5 8B-A1B rose from 28 to 56. Two models moved as much or more and lost on the rule: Qwen3 Coder 30B gained 52 points but rates TIGHT on 24 GB at 8K, and Mistral Small 3.2 gained the same 28 points but is on cooldown until week 43 as pick number three. The gain came from downloads rather than trending: 30-day downloads went from 26,264 to 32,113, up 22 percent in seven days, while the Hub trending score moved from 6 to 9.
The model is a mixture of experts with 8.5 billion parameters in total; its name says about one billion are active per token, a figure we take from the maker and do not measure. What we measure is the file: the Q4_K_M weights weigh 4.8 GB. With an 8K context the engine computes 5.9 GB in total, GOOD on the RTX 3060 Ti, 3070, 4060 and 5060 with their 8 GB, and EXCELLENT on everything from 10 GB upward. At 32K context the total is 6.9 GB: TIGHT on the 8 GB cards, GOOD on the RTX 3080 and RTX 2080 Ti, EXCELLENT from 12 GB. The model accepts up to 128K tokens. All 32 hardware profiles we track fit it at 8K, the Mac mini M6 16GB included.
Renting is the wrong tool for a 6 GB model, and this week's numbers say so plainly. The pick card froze an RTX 4090 hour on Vast.ai at $1.07, the 18:00 UTC hourly median across 50 offers; the day's median was $0.67 and Monday's $0.53. The card you would rent has four times the memory this model needs. An 8 GB card that is already in your machine runs it at 8K context, a 12 GB card runs it at 32K with room to spare, and the break-even question never arises.
- Maker
- Liquid AI
- Parameters
- 8.5B MoE
- Measured GGUF
- Q4_K_M · 4.8 GB
- Max context
- 125K tokens
- Curated for
- general, agents
- Repo created
- 2026-05-28
- Trending score
- 9
- 7-day downloads
- +22.3%
Model pageHugging Face repo (External link)GGUF file (External link)
Downloads, daily readingsdaily last · UTC
2026-10-03 · 32K2026-10-16 · 32K
Watch out
- The rent figure on this card is the 18:00 UTC hourly Vast.ai median for the RTX 4090 (50 offers); the day's median on 16 October was $0.67 and Monday's $0.53. Hourly medians move with which offers happen to be listed.
- Heat Score measures attention on the Hub (downloads, trending, likes), not quality. We publish no benchmarks and no speed figures for this or any model.
- The active-parameter count comes from the maker's model name and card; we measure file size and memory, not throughput.
- Fit verdicts assume llama.cpp-style inference with the measured Q4_K_M GGUF; other runtimes and quantizations change the numbers.
How we chose it
Pick of the week is the model with the largest Heat Score rise over the last seven days among tracked models that run comfortably (GOOD or EXCELLENT) on at least one consumer card of up to 24 GB at 8K context. A rise counts only when it is at least three points, the model's Heat Score is at least 50 today, and the model's own direction is not falling — a low score that climbs while the models around it fall has only moved because they moved. Ties break by the higher Heat Score, then by 30-day downloads. A model cannot be picked again within four weeks. While the heat history is younger than a week, the rise is measured over the readings that exist and the report says so; when no rise counts, the rule falls back to the highest Heat Score and records that.
This issue was selected on the heat rise. Rise measured since 2026-10-10 18:15 UTC. 46 scored models were considered; the full list with reasons is stored with the issue.
The rule on the methodology page · rule v1.2 · selected 2026-10-16 18:16 UTC
Issues
- Week 42, 2026 · LFM2.5 8B-A1B2026-10-16
- Week 41, 2026 · Ornith 1.5 9B2026-10-09
- Week 40, 2026 · Granite 4.2 8B2026-10-02
- Week 39, 2026 · Mistral Small 3.22026-09-25
- Week 38, 2026 · Gemma 4 26B-A4B2026-09-18
- Week 37, 2026 · Gemma 4 12B2026-09-11
A new pick every Friday. The model is computed, the text is written and released by a person, and every number links back to where it was measured.