- Services
A research method on a frozen Qwen3-4B-Instruct-2507 that answers closed-choice fields by reading the scores of option letters instead of generating JSON. Its preregistered study found the same accuracy as grammar-constrained JSON with no decoding cost. It covers Choice and yes/no questions with up to 26 options.
A situational 4.02B-parameter dense decision model from r-ms. Treat the modality benchmarks above as the leading indicator of fit — composite scoring across modalities is still maturing. Newly released, so production-readiness is still being shaken out.
Generated from this model’s benchmarks and ranking signals. Editor reviews refine it over time.
The top devices for this model at 4-bit, ranked by fit and speed.
| Device | Grade | VRAM |
|---|---|---|
| ACEMAGIC M1A Pro (i9-13900HK + ARC A770)ACEMAGIC | SS | 3.0 GB |
| Acer Veriton GN100 AI MiniAcer | SS | 3.0 GB |
| AMD Instinct MI300XAMD | SS | 3.0 GB |
| AMD Instinct MI325XAMD | SS | 3.0 GB |
| AMD Instinct MI355XAMD | SS | 3.0 GB |
Cheapest current cloud rentals with at least 3 GB VRAM, refreshed hourly.
Advertising disclosure: we earn commissions when you shop through the links below.
| Option | Cost / GPU-hour |
|---|---|
NVIDIA GeForce RTX 4060Vast.ai · Spot · 8 GB VRAM | $0.03 |
NVIDIA GeForce RTX 2080 TiVast.ai · Spot · 11 GB VRAM | $0.04 |
NVIDIA GeForce RTX 3070Vast.ai · Spot · 8 GB VRAM | $0.05 |
NVIDIA GeForce RTX 3060Vast.ai · Spot · 12 GB VRAM | $0.05 |
NVIDIA GeForce RTX 4060 TiVast.ai · Spot · 8 GB VRAM | $0.05 |
Per-GPU rate across RunPod, the Vast.ai marketplace, DigitalOcean, and Vultr.
Spot tier is interruptible. Plan for restarts when comparing against on-demand prices.
Explore the Provider
Aggregate stats, leaderboard, release timeline, and benchmark coverage across every r-ms model we track.
Check which of your devices can run this model and how fast.
Compare hosted per-token prices before you commit to local hardware.
Break-even math for self-hosting, renting a GPU, or paying per token.
Live status for the hosted APIs you might use instead of running this locally.
