Underdog Saluki 27B 1.0
consumerA 54GB model in one 8GB file, with agent skills intact. Underdog Saluki 27B 1.0 (Conway Research, October 8) rebuilds Qwen3.8-27B as a tuned 2-bit mix: one 7.89GB GGUF that stock llama.cpp serves, with tool calling scored at 88 against the full model’s 84 on the team’s own bench, parallel tool calls at 42 against 35, and 96 percent average knowledge retention across 9 tests.
The trade is tuned for agents, not leaderboards. Most 2-bit quants chase benchmark numbers and quietly lose the function-call format that turns a chat model into an agent. Saluki’s compression is built the other way around: the agentic behaviors are what it is tuned to keep, and the vision cases are covered by an optional mmproj file (928MB at F16) rather than baked into the main file.
Who it serves. An 8GB machine, a plain mini PC, any card from the last decade: the whole 27B class runs without a rental. Apache 2.0, thinking mode on by default and switchable per request.
- 27.0B
- 262k
- apache 2.0
- Oct 2026
Save your hardware and every model page answers the real question: will it run on your machine, and how fast?
Join free - save your rig →Run it locally
Per-quant memory needs and a static "can you run it?" reference - no rig entry required
The reference hardware
22 reference configs, drawn in-house. Scroll for more.
Can you run it? - reference rigs
| Rig | Q2_K |
|---|---|
| Single GTX 1080 Ti (11GB) | tight |
| 4x H100 80GB (320GB) | fast 896.6t/s |
| NVIDIA DGX Station 748GB | fast 535.3t/s |
| 8x RTX 3090 rack (192GB) | fast 501.1t/s |
| 4x RTX 5090 (128GB) | fast 479.6t/s |
| AMD Instinct MI300X (192GB) | fast 356.3t/s |
| 4x RTX 4090 (96GB) | fast 269.8t/s |
| 2x RTX 5090 (64GB) | fast 239.8t/s |
| 2x RTX 3090 (48GB) | fast 125.3t/s |
| Single RTX 5090 (32GB) | fast 119.9t/s |
| RTX PRO 6000 Blackwell (96GB) | fast 119.9t/s |
| Mac Studio M4 Ultra 192GB | fast 79.7t/s |
| Mac Studio M4 Ultra 512GB | fast 79.7t/s |
| Single RTX 4090 (24GB) | fast 67.5t/s |
| MacBook Pro M5 Max 128GB | fast 44.8t/s |
| Dual EPYC 9004 + 768GB DDR5-4800 | fast 30.8t/s |
| DGX Spark 128GB unified | ok 18.3t/s |
| Ryzen AI Max+ 395 128GB | ok 17.1t/s |
| Jetson AGX Orin 64GB | ok 13.7t/s |
| Epyc + 512GB DDR4-3200 + 2x RTX 3090 | ok 13.7t/s |
| Epyc + 512GB DDR4-2400 + 2x RTX 3090 | ok 10.3t/s |
| NVIDIA Jetson Orin NX 16GB | slow 6.9t/s |
Fit tiers use the same will-it-run logic as the rig finder. For comfortable fits, the badge reflects decode speed: fast >=20 t/s, ok 8-20 t/s, slow <8 t/s. t/s is a bandwidth estimate, not a measured benchmark.
Download options
Or run it in the cloud
No per-token API provider pricing tracked for Underdog Saluki 27B 1.0 yet. For flagship list prices, see the calculator.
Inference cost over time
Data accumulates from the first daily sync - longer ranges populate over time. Prices come from OpenRouter snapshots, not a historical API.
Tokenstead community
Need a second opinion on this model?
Ask the people running it on real hardware, then post your tok/s and fit notes so the next reader has a data point.
Something wrong or stale? Send a correction.