Mistral-Small-24B-Instruct-2501 on RTX 4060 Ti: Tight fit
Conditional memory judgement for one configuration. Change context, quant, or engine and the conclusion can change. Configuration used on this page: GGUF Q4_K_M (Unsloth), 8,192 cached tokens, 1 sequence, engine llama.cpp. Status: Tight. Software: Not verified..
Verdict on GeForce RTX 4060 Ti 16GB (desktop)
Only just: Q4_K_M is the one file that gets in. It needs 14.7 GiB–15.5 GiB at 8,192 tokens against 15.0 GiB of usable VRAM (after the driver reserve). Close other GPU apps first, and expect to cut context if the upper estimate is hit.
With Q4_K_M, the longest context that still fits at all (tight, no headroom) is about 9,997 tokens of a 32,768-token native window, one sequence, fp16 KV cache.
| GGUF quant | File weights | KV cache | Estimated total | Verdict |
|---|---|---|---|---|
| Q8_0 | 23.3 GiB | 1.3 GiB | 24.7 GiB–25.6 GiB | Does not fit |
| Q6_K | 18.0 GiB | 1.3 GiB | 19.4 GiB–20.2 GiB | Does not fit |
| Q5_K_M | 15.6 GiB | 1.3 GiB | 17.0 GiB–17.8 GiB | Does not fit |
| Q4_K_M (recommended) | 13.3 GiB | 1.3 GiB | 14.7 GiB–15.5 GiB | Tight |
Engine llama.cpp, one sequence, fp16 KV. File weights are the published GGUF sizes in the catalog. Totals come from the same calculator that powers the tool below; they are estimates, not a measurement of this setup.
Fixed page configuration
GGUF Q4_K_M needs 14.7 GiB-15.5 GiB at 8,192 tokens against 15.00 GiB net on GeForce RTX 4060 Ti 16GB (desktop), a tight fit. Q8_0, Q6_K, Q5_K_M are tight or over budget.
- ModelMistral-Small-24B-Instruct-2501
- HardwareGeForce RTX 4060 Ti 16GB (desktop)
- QuantizationGGUF Q4_K_M (Unsloth)
- Context length (cached tokens)8,192
- Engine
llama.cpp - Concurrent sequences1
- KV cache typefp16
Technical identifiers and share data
- Model
mistral-small-24b-instruct-2501 - Hardware
nvidia-geforce-rtx-4060-ti-16gb-desktop - Quantization
gguf-q4_k_m - Share fragment for this identity
v2.eyJzY2hlbWFWZXJzaW9uIjoyLCJjYWxjdWxhdG9yVmVyc2lvbiI6IjAuMy4wIiwiZGF0YXNldFZlcnNpb24iOiIyMDI2LTEwLTAyLjEiLCJjb25maWciOnsibW9kZWxJZCI6Im1pc3RyYWwtc21hbGwtMjRiLWluc3RydWN0LTI1MDEiLCJoYXJkd2FyZUlkIjoibnZpZGlhLWdlZm9yY2UtcnR4LTQwNjAtdGktMTZnYi1kZXNrdG9wIiwiZGV2aWNlQ291bnQiOjEsInF1YW50aXphdGlvbklkIjoiZ2d1Zi1xNF9rX20iLCJjb250ZXh0VG9rZW5zIjo4MTkyLCJjb25jdXJyZW50U2VxdWVuY2VzIjoxLCJlbmdpbmUiOiJsbGFtYS5jcHAiLCJrdlR5cGUiOiJmcDE2In19
Memory breakdown
- Weight artifact
- 13.35 GiB (14,333,909,248 bytes)
- Raw KV cache
- 1.25 GiB (1,342,177,280 bytes)
- Engine extras / workspace
- 0.13 GiB–0.85 GiB
- Total estimated demand
- 14.72 GiB–15.45 GiB
- Net calculator budget
- 15.00 GiB
- OS / display reserve
- 1.00 GiB
- Remaining vs net budget
- -0.45 GiB–0.28 GiB
15.45 GiB high estimate vs 15.00 GiB net
Enough memory does not mean this engine, quant, and OS will run. Support does not mean it will be fast.
Other published files on this device
Same model, device and context (8,192 tokens, 1 sequence, FP16 KV). Each row is one published weight-file set with its usual engine, sorted by memory.
| Quantization | Engine | Estimated memory | Fit |
|---|---|---|---|
| BF16 safetensors | vllm | 45.78 GiB–47.69 GiB | Does not fit |
| GGUF Q8_0 (Unsloth) | llama.cpp | 24.71 GiB–25.64 GiB | Does not fit |
| GGUF Q6_K (Unsloth) | llama.cpp | 19.39 GiB–20.22 GiB | Does not fit |
| GGUF Q5_K_M (bartowski) | llama.cpp | 16.99 GiB–17.76 GiB | Does not fit |
| GGUF Q4_K_M (Unsloth) This page | llama.cpp | 14.72 GiB–15.45 GiB | Tight fit |
Assumptions that affect this result
- Requested context is within the recorded native window of 32768 tokens.
- Full-resident weight estimates use total parameters, including inactive MoE experts.
- Weight bytes come from the selected weight files only, not from every file in the repository.
- Download file size is not peak GPU memory. Runtime layout, allocator padding, and KV are extra.
- Counted files: Mistral-Small-24B-Instruct-2501-Q4_K_M.gguf.
- Context budget is the number of cached tokens per sequence (prompt plus reserved generation).
- Concurrent sequences share one model replica; weights are not multiplied by sequence count.
- KV dtype is independent of weight quantization unless you change it.
Related pages
Sources
- Mistral-Small-24B-Instruct-2501 model card
- config.json — revision 9527884be6e5616bdd54de542f9ae13384489724
- safetensors index — metadata.total_size=47144806400
- Artifact repo unsloth/Mistral-Small-24B-Instruct-2501-GGUF
- Artifact repo bartowski/Mistral-Small-24B-Instruct-2501-GGUF
- NVIDIA GeForce RTX 40 Series specs — RTX 4060 Ti Memory Size 16 GB / 8 GB — this record is the 16 GB SKU only
Calculator
This combo page keeps the route model and device. Customize to change them on the main calculator; the share then keeps that identity.
Loading catalog snapshot…
Search or pick a catalog checkpoint to see memory demand.