LFM2.5-24B-A2B Uncensored has 23.972 billion total parameters. The selected Q4_K_M GGUF file is 15.36 GB.
Data snapshot: . 1 published test in this catalog.
Model file and details
Provider
AtomicChat
Total parameters
23,971,587,712
Exact file size
15,359,771,360 bytes
Quantization / precision
Q4_K_M
Weight repository
AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF
Original model repository
AtomicChat/LFM2.5-24B-A2B-Uncensored-GGUF
Weight revision
1acb8f6eb027d85b7771dcc9eb68817aa480e9c8
Model revision
1acb8f6eb027d85b7771dcc9eb68817aa480e9c8
Runtime notes
Confirm model architecture and quantization support in your runner.
Recorded release date
2026-10-09
Date evidence
AtomicChat created this derivative repository on 9 October 2026 and pinned the selected revision after its 13 October update. The publisher documents a refusal-direction edit applied only to blocks 14-39, an exact 15,359,771,360-byte Q4_K_M artifact, llama.cpp revision and H100 evaluation conditions. The selected quant was measured for refusal behavior and divergence from the original, while the reported MMLU result is for the BF16 edited weights rather than this quantized file.
These scores were compiled from the linked sources. These published scores were not independently tested here. Different prompts, precision, thinking budgets and tools can change results.
MMLU · AtomicChat 2,000-question sample
65.55 / 100
AtomicChat publisher evaluation on a fixed 2,000-question MMLU sample used while selecting this weight edit. Kept separate from full MMLU and other sampled protocols.
Benchmark efficiency: 4.268 published score points per GB of selected weights. This ratio is not an inference-speed measure; check the test settings to see whether the score covers the selected quantization.
Test settings: AtomicChat publisher evaluation of the edited BF16 weights on a fixed 2,000-question MMLU sample using 1x H100 PCIe, llama.cpp revision 980aef8c and source weights revision f772950c. The original scored 64.40% and the edited model 65.55%; McNemar p=0.06.
Publisher-reported, not an independent test. The 2,000-question set was part of edit selection, so this is not a held-out quality result and is kept separate from full MMLU. AtomicChat did not separately report MMLU for the selected Q4_K_M artifact. It did report that exact quant at 15.4 GB, mean KLD 0.274 and 79.6% same top-1 versus the original BF16 model, plus 1.2%/0%/0% refusal rates on its English/Russian/XSTest sets. Refusal detection inspects reply openings and can miss soft refusals. Removing learned refusals does not make outputs safe, correct or lawful. The 15,359,771,360-byte file excludes runtime and KV-cache headroom; a 24 GB machine recommendation is publisher guidance, not proof of compatibility or speed for every configuration. The LFM Open License v1.0 applies.
Context and version evidence: The repository and file name identify the selected model. Revision hashes are shown above when recorded. Tested context length, full hardware configuration and run date are not consistently recorded in this snapshot. Only settings explicitly stated in each benchmark record are known.
The 15.36 GB figure covers model weights in this file. It excludes the context cache, runtime, operating system and any separate multimodal files. It is not a measured RAM requirement or a token-generation speed.
These complete configurations pass the initial weight-size check for the selected 15.36 GB file. Allow extra memory for context and software. Check the model’s runtime notes and your inference software before buying.
Listing evidence: 2026-09-25. Amazon.com base price visible in an owner-supplied screenshot reviewed 25 Sep 2026; capture time not supplied. Dated reference, not a live quote. Excludes card signup offers, coupons, tax and shipping.
Listing evidence: 2026-10-03. Apple US launch price announced 22 September 2026: USD 1,699. Dated manufacturer reference, not an Amazon offer; tax and shipping may differ.
Listing evidence: 2026-09-24. GMKtec, EVO-X2 / 64GB / 1TB, recorded 22 Sep 2026; exact ASIN capacity identified 24 Sep. This is a dated reference, not a current Amazon price; tax, shipping and bundle contents may differ.
Memory matches are estimates, not tested compatibility or speed. As an Amazon Associate I earn from qualifying purchases.
RAM and VRAM planning for LFM2.5 24B-A2B Uncensored
Confirm model architecture and quantization support in your runner. The selected weights use 15.36 GB; this excludes context cache, runtime buffers and other programs.
These are the first listed tiers that pass our reserve policy, not measured minimum requirements. System RAM and VRAM remain separate. Use the memory tool to change the reserve for your context and runner.