LFM2.5 2.6B hardware requirements
What is LFM2.5 2.6B?
LFM2.5 2.6B is a small Liquid AI model released in August 2026 for on-device agents, mixing short convolution blocks with attention layers. It is text-only with a 128k context and 16 languages including Korean, and Liquid AI recommends it for tool use, data extraction and RAG but not agentic coding. It always thinks before answering, and its LFM Open License 1.0 excludes commercial use by companies with annual revenue of 10 million US dollars or more.
About the model
- Parameters
- 2.7B
- Attention
- Hybrid attention (only some layers use full attention)
- Max context
- 131,072 tokens
- Released
- 2026-07-28
- Model card
- LFM2.5 2.6B on Hugging Face
GGUF files
| Quant | File size | Quality | Published by |
|---|---|---|---|
| Q4_K_Mbaseline | 1.67 GB | Good balance | Official GGUF · LiquidAI |
| Q8_0 | 2.87 GB | Near-lossless | Official GGUF · LiquidAI |
Memory needed by quant and context
| Quant | 4k | 8k | 16k | 32k | 64k | 128k |
|---|---|---|---|---|---|---|
| Q4_K_M | 2.3 | 2.4 | 2.6 | 3.0 | 3.8 | 5.5 |
| Q8_0 | 3.5 | 3.6 | 3.8 | 4.2 | 5.0 | 6.7 |
Which GPUs can run LFM2.5 2.6B?
How to read the verdicts
- Runs great
- Fully on the GPU at 20 tok/s or more
- Runs well
- Fully on the GPU at 8–20 tok/s; MoE experts in system RAM at 20 tok/s or more; or at least 90% on the GPU at 8 tok/s or more
- Runs slowly
- 2–8 tok/s; CPU-only; less than 90% on the GPU; or MoE experts in system RAM below 20 tok/s
- Won't run
- Does not fit, or under 2 tok/s
Filter this list by budget, type and context in the calculator
Scroll sideways to see every column.
| Hardware | Verdict | Speed | Quant | Memory | Runs as | Context | Notes | Price |
|---|---|---|---|---|---|---|---|---|
| GeForce RTX 5090 | Runs great | est. 695.3 tok/s486.7–903.8theoretical estimate ±30% | Q4_K_M | 3.0 / 32.0 GB | Full GPU | up to 128k | — | $4,200 street (as of 2026-08-09) |
| Radeon RX 7900 XTX | Runs great | est. 404.4 tok/s283.1–525.7theoretical estimate ±30% | Q4_K_M | 3.0 / 24.0 GB | Full GPU | up to 128k | — | $999 launch MSRP |
| GeForce RTX 3090 Ti | Runs great | est. 391.1 tok/s273.8–508.4theoretical estimate ±30% | Q4_K_M | 3.0 / 24.0 GB | Full GPU | up to 128k | — | $1,999 launch MSRP |
| GeForce RTX 4090 | Runs great | est. 391.1 tok/s273.8–508.4theoretical estimate ±30% | Q4_K_M | 3.0 / 24.0 GB | Full GPU | up to 128k | — | $2,755 street (as of 2026-08-09) |
| Apple M5 Ultra · 96 GB | Runs great | est. 388.0 tok/s271.6–504.3theoretical estimate ±30% | Q4_K_M | 2.4 / 67.2 GB | Unified memory | up to 128k | — | — |
| GeForce RTX 5080 | Runs great | est. 372.5 tok/s260.7–484.2theoretical estimate ±30% | Q4_K_M | 3.0 / 16.0 GB | Full GPU | up to 128k | — | $1,256 street (as of 2026-08-09) |
| GeForce RTX 3090 | Runs great | est. 363.1 tok/s254.2–472.1theoretical estimate ±30% | Q4_K_M | 3.0 / 24.0 GB | Full GPU | up to 128k | — | $1,050 street (as of 2026-08-09) |
| GeForce RTX 3080 12GB | Runs great | est. 353.8 tok/s247.7–460.0theoretical estimate ±30% | Q4_K_M | 3.0 / 12.0 GB | Full GPU | up to 128k | — | $799 launch MSRP |
| GeForce RTX 5070 Ti | Runs great | est. 347.6 tok/s243.3–451.9theoretical estimate ±30% | Q4_K_M | 3.0 / 16.0 GB | Full GPU | up to 128k | — | $949 street (as of 2026-08-09) |
| GeForce RTX 5080 Laptop | Runs great | est. 347.6 tok/s243.3–451.9theoretical estimate ±30% | Q4_K_M | 3.0 / 16.0 GB | Full GPU | up to 128k | — | — |
| GeForce RTX 5090 Laptop | Runs great | est. 347.6 tok/s243.3–451.9theoretical estimate ±30% | Q4_K_M | 3.0 / 24.0 GB | Full GPU | up to 128k | — | — |
| Radeon RX 7900 XT | Runs great | est. 337.0 tok/s235.9–438.1theoretical estimate ±30% | Q4_K_M | 3.0 / 20.0 GB | Full GPU | up to 128k | — | $899 launch MSRP |
| GeForce RTX 3080 10GB | Runs great | est. 294.9 tok/s206.4–383.3theoretical estimate ±30% | Q4_K_M | 3.0 / 10.0 GB | Full GPU | up to 128k | — | $699 launch MSRP |
| GeForce RTX 4080 Super | Runs great | est. 285.6 tok/s199.9–371.2theoretical estimate ±30% | Q4_K_M | 3.0 / 16.0 GB | Full GPU | up to 128k | — | $999 launch MSRP |
| GeForce RTX 4080 | Runs great | est. 278.2 tok/s194.7–361.6theoretical estimate ±30% | Q4_K_M | 3.0 / 16.0 GB | Full GPU | up to 128k | — | $1,199 launch MSRP |
| Radeon RX 9070 | Runs great | est. 271.7 tok/s190.2–353.2theoretical estimate ±30% | Q4_K_M | 3.0 / 16.0 GB | Full GPU | up to 128k | — | $639 street (as of 2026-08-09) |
| Radeon RX 9070 XT | Runs great | est. 271.7 tok/s190.2–353.2theoretical estimate ±30% | Q4_K_M | 3.0 / 16.0 GB | Full GPU | up to 128k | — | $689 street (as of 2026-08-09) |
| Radeon RX 7800 XT | Runs great | est. 262.9 tok/s184.0–341.7theoretical estimate ±30% | Q4_K_M | 3.0 / 16.0 GB | Full GPU | up to 128k | — | $499 launch MSRP |
| GeForce RTX 4070 Ti Super | Runs great | est. 260.7 tok/s182.5–338.9theoretical estimate ±30% | Q4_K_M | 3.0 / 16.0 GB | Full GPU | up to 128k | — | $799 launch MSRP |
| GeForce RTX 5070 | Runs great | est. 260.7 tok/s182.5–338.9theoretical estimate ±30% | Q4_K_M | 3.0 / 12.0 GB | Full GPU | up to 128k | — | $629 street (as of 2026-08-09) |
| GeForce RTX 5070 Ti Laptop | Runs great | est. 260.7 tok/s182.5–338.9theoretical estimate ±30% | Q4_K_M | 3.0 / 12.0 GB | Full GPU | up to 128k | — | — |
| Apple M3 Ultra · 96 GB | Runs great | est. 258.7 tok/s181.1–336.4theoretical estimate ±30% | Q4_K_M | 2.4 / 67.2 GB | Unified memory | up to 128k | — | — |
| Apple M5 Max (40-core GPU) · 48 GB | Runs great | est. 258.6 tok/s181.0–336.2theoretical estimate ±30% | Q4_K_M | 2.4 / 33.6 GB | Unified memory | up to 128k | — | — |
| Apple M2 Ultra · 64 GB | Runs great | est. 252.7 tok/s176.9–328.6theoretical estimate ±30% | Q4_K_M | 2.4 / 44.8 GB | Unified memory | up to 128k | — | — |
| GeForce RTX 2080 Ti | Runs great | est. 239.0 tok/s167.3–310.7theoretical estimate ±30% | Q4_K_M | 3.0 / 11.0 GB | Full GPU | up to 128k | — | $999 launch MSRP |
| Apple M4 Max (40-core GPU) · 48 GB | Runs great | est. 230.0 tok/s161.0–299.0theoretical estimate ±30% | Q4_K_M | 2.4 / 33.6 GB | Unified memory | up to 128k | — | — |
| GeForce RTX 4090 Laptop | Runs great | est. 223.5 tok/s156.4–290.5theoretical estimate ±30% | Q4_K_M | 3.0 / 16.0 GB | Full GPU | up to 128k | — | — |
| GeForce RTX 4070 | Runs great | est. 195.5 tok/s136.9–254.2theoretical estimate ±30% | Q4_K_M | 3.0 / 12.0 GB | Full GPU | up to 128k | — | $599 launch MSRP |
| GeForce RTX 4070 Super | Runs great | est. 195.5 tok/s136.9–254.2theoretical estimate ±30% | Q4_K_M | 3.0 / 12.0 GB | Full GPU | up to 128k | — | $599 launch MSRP |
| Apple M5 Max (32-core GPU) · 36 GB | Runs great | est. 193.8 tok/s135.6–251.9theoretical estimate ±30% | Q4_K_M | 2.4 / 25.2 GB | Unified memory | up to 128k | — | — |
| GeForce GTX 1080 Ti | Runs great | est. 187.8 tok/s131.4–244.1theoretical estimate ±30% | Q4_K_M | 3.0 / 11.0 GB | Full GPU | up to 128k | — | $699 launch MSRP |
| GeForce RTX 3070 | Runs great | est. 173.8 tok/s121.7–226.0theoretical estimate ±30% | Q4_K_M | 3.0 / 8.0 GB | Full GPU | up to 128k | — | $499 launch MSRP |
| GeForce RTX 5060 | Runs great | est. 173.8 tok/s121.7–226.0theoretical estimate ±30% | Q4_K_M | 3.0 / 8.0 GB | Full GPU | up to 128k | — | $339 street (as of 2026-08-09) |
| GeForce RTX 5060 Ti 16GB | Runs great | est. 173.8 tok/s121.7–226.0theoretical estimate ±30% | Q4_K_M | 3.0 / 16.0 GB | Full GPU | up to 128k | — | $569 street (as of 2026-08-09) |
| GeForce RTX 5060 Ti 8GB | Runs great | est. 173.8 tok/s121.7–226.0theoretical estimate ±30% | Q4_K_M | 3.0 / 8.0 GB | Full GPU | up to 128k | — | $429 street (as of 2026-08-09) |
| Apple M4 Max (32-core GPU) · 36 GB | Runs great | est. 172.7 tok/s120.9–224.5theoretical estimate ±30% | Q4_K_M | 2.4 / 25.2 GB | Unified memory | up to 128k | — | — |
| GeForce RTX 4080 Laptop | Runs great | est. 167.6 tok/s117.3–217.9theoretical estimate ±30% | Q4_K_M | 3.0 / 12.0 GB | Full GPU | up to 128k | — | — |
| Intel Arc B580 | Runs great | est. 149.1 tok/s104.4–193.9theoretical estimate ±30% | Q4_K_M | 3.0 / 12.0 GB | Full GPU | up to 128k | — | $290 street (as of 2026-08-09) |
| GeForce RTX 3060 12GB | Runs great | est. 139.7 tok/s97.8–181.6theoretical estimate ±30% | Q4_K_M | 3.0 / 12.0 GB | Full GPU | up to 128k | — | $250 street (as of 2026-08-09) |
| Radeon RX 9060 XT 16GB | Runs great | est. 135.6 tok/s94.9–176.3theoretical estimate ±30% | Q4_K_M | 3.0 / 16.0 GB | Full GPU | up to 128k | — | $449 street (as of 2026-08-09) |
| Apple M5 Pro · 24 GB | Runs great | est. 129.3 tok/s90.5–168.1theoretical estimate ±30% | Q4_K_M | 2.4 / 16.8 GB | Unified memory | up to 128k | — | — |
| NVIDIA DGX Spark · 64 GB | Runs great | est. 118.0 tok/s82.6–153.4theoretical estimate ±30% | Q4_K_M | 2.4 / 48.2 GB | Unified memory | up to 128k | — | $3,999 launch MSRP (128 GB) |
| Ryzen AI Max+ 395 (Strix Halo) · 32 GB | Runs great | est. 116.3 tok/s81.4–151.3theoretical estimate ±30% | Q4_K_M | 2.4 / 22.4 GB | Unified memory | up to 128k | — | $1,999 launch MSRP (128 GB) |
| Apple M4 Pro · 24 GB | Runs great | est. 115.0 tok/s80.5–149.5theoretical estimate ±30% | Q4_K_M | 2.4 / 16.8 GB | Unified memory | up to 128k | — | — |
| GeForce RTX 4060 Ti 16GB | Runs great | est. 111.7 tok/s78.2–145.3theoretical estimate ±30% | Q4_K_M | 3.0 / 16.0 GB | Full GPU | up to 128k | — | $499 launch MSRP |
| GeForce RTX 4060 Ti 8GB | Runs great | est. 111.7 tok/s78.2–145.3theoretical estimate ±30% | Q4_K_M | 3.0 / 8.0 GB | Full GPU | up to 128k | — | $399 launch MSRP |
| GeForce RTX 4060 | Runs great | est. 105.5 tok/s73.9–137.2theoretical estimate ±30% | Q4_K_M | 3.0 / 8.0 GB | Full GPU | up to 128k | — | $299 launch MSRP |
| GeForce RTX 4060 Laptop | Runs great | est. 99.3 tok/s69.5–129.1theoretical estimate ±30% | Q4_K_M | 3.0 / 8.0 GB | Full GPU | up to 128k | — | — |
| GeForce RTX 4070 Laptop | Runs great | est. 99.3 tok/s69.5–129.1theoretical estimate ±30% | Q4_K_M | 3.0 / 8.0 GB | Full GPU | up to 128k | — | — |
| GeForce RTX 3050 8GB | Runs great | est. 86.9 tok/s60.8–113.0theoretical estimate ±30% | Q4_K_M | 3.0 / 8.0 GB | Full GPU | up to 128k | — | $249 launch MSRP |
| Apple M5 · 16 GB | Runs great | est. 64.7 tok/s45.3–84.1theoretical estimate ±30% | Q4_K_M | 2.4 / 11.2 GB | Unified memory | up to 128k | — | — |
| Apple M6 · 16 GB | Runs great | est. 64.7 tok/s45.3–84.1theoretical estimate ±30% | Q4_K_M | 2.4 / 11.2 GB | Unified memory | up to 128k | — | — |
| Apple M4 · 16 GB | Runs great | est. 50.5 tok/s35.4–65.7theoretical estimate ±30% | Q4_K_M | 2.4 / 11.2 GB | Unified memory | up to 64k | — | — |
| DDR5-6000 dual-channel CPU · 16 GB | Runs slowly | est. 26.6 tok/s18.6–34.6theoretical estimate ±30% | Q4_K_M | 1.8 GB RAM | CPU only | up to 128k |
| — |
| DDR5-5600 dual-channel CPU · 16 GB | Runs slowly | est. 24.8 tok/s17.4–32.3theoretical estimate ±30% | Q4_K_M | 1.8 GB RAM | CPU only | up to 128k |
| — |
| DDR4-3200 dual-channel CPU · 16 GB | Runs slowly | est. 14.2 tok/s9.9–18.4theoretical estimate ±30% | Q4_K_M | 1.8 GB RAM | CPU only | up to 128k |
| — |
Models with a thinking mode spend extra tokens thinking, so their speed thresholds are 1.5× stricter (30 / 12 / 3 tok/s). With thinking turned off, a model may do one grade better.
Cheapest GPUs that run it
- Runs greatGeForce RTX 3060 12GB$250 street (as of 2026-08-09), est. 139.7 tok/s (97.8–181.6, theoretical estimate ±30%).
Measured results for LFM2.5 2.6B
No public measurements for this model yet.
Frequently asked questions
How much VRAM does LFM2.5 2.6B need?
At Q4_K_M the file is 1.67 GB; with 8k context a graphics card needs about 2.4 GB (weights 1.66 GB, KV cache 0.1 GB, compute buffer 0.6 GB).
What is the cheapest GPU that runs LFM2.5 2.6B well?
Cheapest that runs it great: GeForce RTX 3060 12GB at $250 (street price as of 2026-08-09), est. 139.7 tok/s (97.8–181.6, theoretical estimate ±30%).
Can LFM2.5 2.6B run on an 8 GB, 16 GB or 24 GB GPU?
At Q4_K_M with 8k context: GeForce RTX 4060 — Runs great, est. 105.5 tok/s (73.9–137.2, theoretical estimate ±30%); GeForce RTX 5060 Ti 16GB — Runs great, est. 173.8 tok/s (121.7–226.0, theoretical estimate ±30%); GeForce RTX 3090 — Runs great, est. 363.1 tok/s (254.2–472.1, theoretical estimate ±30%).
Speeds are estimates from memory bandwidth, calibrated against public benchmarks, and each one comes with an error band and a confidence label. Real results vary with drivers, backend, context length and thermals.