StarCoder2 15B · 11 GPUs measured first-party · llama.cpp Q4_K_M · Updated October 2026

What GPU Do You Need for StarCoder2 15B?

StarCoder2 15B is the BigCode project's open code model, trained on permissively-licensed, documented source data with fill-in-the-middle as a first-class objective. We measured it on 11 GPUs (llama.cpp, Q4_K_M): 144 tok/s on the B300, ~11GB peak VRAM.

Benchmarked weights: bartowski/starcoder2-15b-instruct-v0.1-GGUF

Fastest we measured

NVIDIA B300

144.1 tok/s on StarCoder2 15B, the ceiling. Measured on our bench. 288GB of VRAM, $40,000 at launch.

Pros

• 144.1 tok/s on StarCoder2 15B
• 288GB, clears the StarCoder2 15B floor
• Rentable by the hour rather than bought

Cons

• 1400W board rating
• Datacenter or workstation hardware, not a retail purchase

Cheapest card that runs it

NVIDIA RTX PRO 6000 Blackwell Workstation Edition

122.1 tok/s on StarCoder2 15B, lowest launch price that still fits. Measured on our bench. 96GB of VRAM, $8,565 at launch.

Pros

• 122.1 tok/s on StarCoder2 15B
• 96GB, clears the StarCoder2 15B floor
• Rentable by the hour rather than bought

Cons

• 600W board rating
• Datacenter or workstation hardware, not a retail purchase

The open-data coder. StarCoder2's distinction mirrors SmolLM3's: the BigCode initiative documents exactly what code it trained on (The Stack v2, license-filtered, with opt-outs honored). For teams that care where their completion model's training code came from, legal departments do, it's the most defensible coder in our lineup. Technically it's a fill-in-the-middle native at a size between our other code picks: bigger than Qwen2.5-Coder 7B, lighter than Codestral 22B.

Where it lands on hardware. The ~11GB floor makes 12GB cards its natural home, an Arc B580 or RTX 3060 12GB hosts it where Codestral's ~14GB wants 16GB. Speed is honest for the size: 144 tok/s peak, 136 on the H200, and a usable 24 tok/s even on an L4. For completions, remember prompt ingestion dominates perceived latency: 2,457-4,428 prompt t/s at the top of the chart means large-file context loads fast.

How it compares. H100 80GB HBM3: StarCoder2 15B 134.9 tok/s, Phi-4 14B (Q3_K_M) 138.9, Gemma 3 12B (Q3_K_M) 125.9, Qwen2.5-14B 144.7 (15B), DeepSeek-R1 Distill 14B 144.8. 3 of 4 beat StarCoder2 15B here.

Cost on a rented GPU. 1M generated tokens of StarCoder2 15B: $1.63 on a A100 40GB SXM4 ($0.47/hr, 3.5 hours), $13.38 on a B300 ($6.94/hr, 116 min, 8.2x the cost).

Speed tiers for StarCoder2 15B. 30+ tok/s: 9 (B300, B200, H200); 10-30 tok/s: 2 (L4, T4). 30 tok/s is roughly where replies outpace reading.

Reading your prompt. Before StarCoder2 15B writes anything it reads the input: 6106.1 tok/s on the RTX PRO 6000 Blackwell Workstation Edition (0.7s for a 4,000-token prompt), 554.6 on the T4 (7.2s). Long documents and big code files feel this number more than the generation speed.

VRAM for StarCoder2 15B. Measured peak 9.8GB, so 12GB is the smallest common card size; smallest card it ran on: T4 (16GB). With long context: Q4_K_M 11GB (tested), Q2_K 7GB, Q3_K_M 9GB, Q5_K_M 13GB, Q6_K 15GB.

Power on StarCoder2 15B. Most efficient: H100 80GB HBM3, 181W, 0.37 kWh per 1M generated tokens. Hungriest: B200, 345W, 0.69 kWh. At $0.15/kWh: $0.056 per 1M generated tokens.

Our verdict

StarCoder2 15B: 144 tok/s peak, ~11GB floor, the documented-training-data coder that fits a 12GB card. Between Qwen's 7B and Codestral's 22B on size, and ahead of both on provenance; if your completion model needs a paper trail, this is it.

FAQ

What GPU does StarCoder2 15B need?

A 12GB card. Measured peak was ~11GB at Q4_K_M. That undercuts Codestral 22B's ~14GB floor by a full card class: B580/RTX 3060 12GB territory instead of 16GB.

Why does its training data matter?

BigCode documents it fully: permissively-licensed code from The Stack v2, license-filtered, developer opt-outs honored. For organizations with IP concerns about AI-generated code, that provenance is the model's headline feature.

StarCoder2 15B or Codestral 22B?

Codestral completes better in our experience. It's the completion specialist pick if 16GB is available. StarCoder2 counters with a lighter floor (12GB cards) and auditable training data. Both are FIM-native; the tradeoff is capability versus footprint-plus-provenance.

Is it fast enough for editor completions?

Yes: completion UX depends on prompt ingestion plus first tokens, and it measured up to 4,428 prompt t/s with 144 tok/s generation. Even mid-tier cloud cards keep suggestions instant-feeling.

What quantization did you test?

Q4_K_M via llama.cpp, the same pinned harness as every LLM here: so its numbers compare directly against Codestral, the Qwen coders and everything else in the database.

How do I pair it in a local coding stack?

Use it as the completion engine next to a bigger agent model: StarCoder2 answers the editor's fill-in-middle requests from a 12GB card while something like Devstral 24B or Qwen3 Coder 30B-A3B handles chat, refactors and agent loops. The split mirrors how the models were trained, and a 24GB card hosts both halves.