Llama 3.2 3B Instruct
edgeAI-generated content marks
The provider reports that this model does not add embedded watermarks or provenance metadata to generated output.
- 3.2B
- 128k
- llama 3
- 🇺🇸 USA
- Sep 2024
Scores
Score per dollar
1000 pts per $/M input
general_score (50) divided by cheapest input price ($0.05/M). Higher is better value. See live pricing.
Related models
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 | Q4_K_M | Q8_0 |
|---|---|---|
| 4x H100 80GB (320GB) | fast 2469.8t/s | fast 1643.6t/s |
| NVIDIA DGX Station 748GB | fast 1474.5t/s | fast 981.3t/s |
| 8x RTX 3090 rack (192GB) | fast 1380.5t/s | fast 918.7t/s |
| 4x RTX 5090 (128GB) | fast 1321.2t/s | fast 879.2t/s |
| AMD Instinct MI300X (192GB) | fast 981.5t/s | fast 653.1t/s |
| 4x RTX 4090 (96GB) | fast 743.2t/s | fast 494.6t/s |
| 2x RTX 5090 (64GB) | fast 660.6t/s | fast 439.6t/s |
| 2x RTX 3090 (48GB) | fast 345.1t/s | fast 229.7t/s |
| Single RTX 5090 (32GB) | fast 330.3t/s | fast 219.8t/s |
| RTX PRO 6000 Blackwell (96GB) | fast 330.3t/s | fast 219.8t/s |
| Mac Studio M4 Ultra 192GB | fast 219.6t/s | fast 146.1t/s |
| Mac Studio M4 Ultra 512GB | fast 219.6t/s | fast 146.1t/s |
| Single RTX 4090 (24GB) | fast 185.8t/s | fast 123.6t/s |
| MacBook Pro M5 Max 128GB | fast 123.5t/s | fast 82.2t/s |
| Single GTX 1080 Ti (11GB) | fast 89.2t/s | fast 59.4t/s |
| Dual EPYC 9004 + 768GB DDR5-4800 | fast 84.9t/s | fast 56.5t/s |
| DGX Spark 128GB unified | fast 50.3t/s | fast 33.5t/s |
| Ryzen AI Max+ 395 128GB | fast 47.2t/s | fast 31.4t/s |
| Jetson AGX Orin 64GB | fast 37.8t/s | fast 25.1t/s |
| Epyc + 512GB DDR4-3200 + 2x RTX 3090 | fast 37.8t/s | fast 25.1t/s |
| Epyc + 512GB DDR4-2400 + 2x RTX 3090 | fast 28.3t/s | ok 18.8t/s |
| NVIDIA Jetson Orin NX 16GB | ok 18.9t/s | ok 12.6t/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
Live per-provider pricing, throughput and uptime - refreshed about 11 hours ago via OpenRouter. Click a column to sort.
| Provider | Type | Input $/M | Output $/M | Cache $/M | Tok/s | Latency | Uptime | Value |
|---|---|---|---|---|---|---|---|---|
|
Parasail
|
API | 0.05 | 0.33 | - | - | - | 100.00% | best uptime |
|
Cloudflare
|
API | 0.05 | 0.34 | - | - | - | 100.00% |
Default order: throughput among 95%+ uptime providers, then latency; subscriptions last. Sort by any column. Subscription rows show $/mo in the Value column - per-token columns are "-". Affiliate links are marked sponsored / nofollow. Confirm current pricing on the provider's site before committing.
Detailed API pricing page + JSON endpoint →
See who runs Meta in production →
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.