Apple · M1 · Desktop
Best AI Models for iMac 24" M1 (16 GB)
Runs 9B models at 8 tok/s; 14B only just fits in 12.5 GB, and 27–31B doesn't fit.
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Q4_K_M · 5.0 GB | 16.8 tok/s | |
| 2 | Q4_K_M · 2.3 GB | 58.3 tok/s | |
| 3 | Gemma 3 1B ITGoogle Q4_K_M · 2.1 GB | 72.4 tok/s | |
| 4 | Ling 3.0 TinyInclusion AI Q4_K_M · 5.5 GB | 41.1 tok/s | |
| 5 | LFM2.5 8B A1BLiquid AI Q4_K_M · 5.6 GB | 36.2 tok/s |
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Q4_K_M · 2.3 GB | 58.3 tok/s | |
| 2 | Ling 3.0 TinyInclusion AI Q4_K_M · 5.5 GB | 41.1 tok/s | |
| 3 | LFM2.5 8B A1BLiquid AI Q4_K_M · 5.6 GB | 36.2 tok/s | |
| 4 | Qwen3.5 2BAlibaba Q4_K_M · 1.9 GB | 27.0 tok/s | |
| 5 | MiniCPM5 2BOpenBMB Q4_K_M · 2.5 GB | 25.2 tok/s |
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Q4_K_M · 5.0 GB | 16.8 tok/s | |
| 2 | Qwen3.5 2BAlibaba Q4_K_M · 1.9 GB | 27.0 tok/s | |
| 3 | Cosmos Reason2 2BNVIDIA Q4_K_M · 3.5 GB | 29.7 tok/s | |
| 4 | LocateAnything 3BNVIDIA Q4_K_M · 3.2 GB | 17.9 tok/s | |
| 5 | Qwen3.5 0.8BAlibaba Q4_K_M · 0.9 GB | 55.0 tok/s |
The iMac 24" M1 is a 16 GB desktop with more memory than speed: it runs 9B models at 8 tok/s. macOS keeps about 3.5 GB, leaving 12.5 GB for models. Prompt reading is its weak spot: a 4K-token prompt waits 34 s before the first word on Gemma 3 4B IT. M1 generation, released 2021; the newer M2 writes 34% faster with 1.4× the compute.
iMac 24" M1 strengths and weaknesses
Weaknesses
- 27–31B models don't fit: Qwen3.8 27B needs 17.9 GB at Q4 vs 12.5 GB usable here
- Little room for context: Qwen3 14B uses 11.9 of 12.5 GB usable, 0.6 GB left
- Slow prompt reading: A 32K-token prompt on Gemma 3 4B IT waits 4.5 min for the first token
- No CUDA: vLLM, AWQ, GPTQ and FP8 builds don't run; use GGUF or MLX
Memory for models
12.5 GB of 16 GB27–31B models need 17.9 GB at Q4, more than the 12.5 GB available.
Other memory sizes of the same machine first, then machines and GPUs people weigh against it, each with the one reason it is worth a look. Speeds on Qwen3.6 35B A3B.
Output speed depends on memory bandwidth (68 GB/s). Grey rows need more than the 12.5 GB available.
- 7.5tok/s
- 5.0tok/s
- 27–31BQwen3.8 27B · does not fitneeds 17.9 GB
- 35B MoEQwen3.6 35B A3B · does not fitneeds 22.0 GB
- 70BLlama 3.3 70B Instruct · does not fitneeds 51.3 GB
- 120B MoEGPT OSS 120B · does not fitneeds 70.8 GB
Estimated speeds: Q4_K_M, 4K context, one user. How we estimate
Chip generations: writing vs compute
M1 → M4Writing speed follows memory bandwidth and barely moved: 24.9 to 38.1 tok/s since the M1. Compute, which sets how fast prompts are read, grew 1.3×. Indexed to the M1; speeds on Qwen3.6 35B A3B, same 16 GB.
All 918 models on the iMac 24" M1
Featured onlyBest quant that fits in 12.5 GB, ordered by llmrun Score. Fit grades how well it runs here, S best. How ranking works
| Model | Quant | VRAM | Speed | Context | Score | Fit |
|---|---|---|---|---|---|---|
Qwen2.5 VL 7B InstructVision | Q4_K_M | 6.2 GB | 8.9 t/s | 128K | B47, Decent | |
TinyLLama V0Chat | Q4_K_M | 0.0 GB | 4781.0 t/s | 2K | S100, Runs great | |
| Q4_K_M | 0.3 GB | 159.4 t/s | 1K | S97, Runs great | ||
| Q4_K_M | 1.7 GB | 55.0 t/s | 33K | A82, Runs well | ||
| Q4_K_M | 0.1 GB | 318.7 t/s | 1K | S100, Runs great | ||
| Q4_K_M | 6.2 GB | 8.9 t/s | 8K | B47, Decent | ||
Ministral 3 14B Reasoning 2512ReasoningFunctions | Q4_K_M | 12.0 GB | 5.3 t/s | 262K | D18, Barely runs | |
Lucid V1 NemoChat | Q4_K_M | 11.0 GB | 5.9 t/s | 131K | C32, Usable | |
| BF16 | 7.1 GB | 7.6 t/s | 8K | B45, Decent | ||
Huihui GPT OSS 20B BF16 AbliteratedChatFunctions | Q3_K_M | 10.8 GB | 20.4 t/s | 131K | B53, Decent | |
| Q3_K_M | 11.1 GB | 18.1 t/s | 131K | B46, Decent | ||
Qwen2.5 VL 3B InstructVision | Q4_K_M | 3.2 GB | 18.2 t/s | 128K | B58, Decent | |
Gemma 3n E4B ITVision | Q4_K_M | 7.7 GB | 9.2 t/s | — | B47, Decent | |
Qwen3 VL 2B ThinkingVisionFunctions | Q4_K_M | 3.5 GB | 26.4 t/s | 262K | A67, Runs well | |
DialoGPT SmallChat | Q4_K_M | 0.1 GB | 398.4 t/s | 1K | S100, Runs great |
Specs
- Chip
- M1
- CPU cores
- 8
- GPU cores
- 8
- Unified memory
- 16 GB
- For models
- 12.5 GB (macOS keeps 3.5 GB)
- Memory bandwidth
- 68.3 GB/s
- FP16 (dense)
- 5.2 TFLOPS
- Neural Engine
- 11 TOPS
- Released
- 2021-05-21
iMac 24" M1 FAQ
How much memory is available for AI on the iMac 24" M1 (16 GB)?
About 12.5 GB of the 16 GB. macOS keeps roughly 3.5 GB for itself and other apps.
What is the largest model the iMac 24" M1 (16 GB) can run?
GPT OSS 20B (20.9B) at Q4_K_S uses 12.4 GB, 99% of the available memory, and runs at about 18.0 tok/s.
Can the iMac 24" M1 (16 GB) run Gemma 3 4B IT?
Yes. Gemma 3 4B IT fits at Q4_K_M, using 5.0 GB of the 12.5 GB available, and it is the best model that runs well on this machine.
Can the iMac 24" M1 (16 GB) run 27–31B models?
No. Qwen3.8 27B needs 17.9 GB at Q4, more than the 12.5 GB available.
How long does the iMac 24" M1 (16 GB) take to read a long prompt?
On Gemma 3 4B IT, about 34 s for a 4K-token prompt and 4.5 min for a 32K-token document before the first word appears.
Is renting a GPU cheaper than buying the iMac 24" M1?
A Tesla V100 (16 GB) rents from $0.10/hr.