Up to date
bigcode BigCode OpenRAIL-M out 3 years ago
16.0B parameters, text in, text out, 16K tokens of context.
Verified 2026-10-25 Rank #759 in LLMs · score 71.3 method v1 Sources 3 Downloads 30d 12.6k · #490 of 619 tracked
Picked by rule from 9 tracked pieces: the lab's own post, the most shared, a field report, the latest.
SelfCodeAlign creates StarCoder2-Instruct by using StarCoder2-15B to generate thousands of instruction-response pairs for fine-tuning StarCoder-15B without human annotations or distilled data from proprietary models.
StarCoder2-15B-Instruct achieves a 72.6 HumanEval score, surpassing the 72.0 score of CodeLlama-70B-Instruct.
The 2 picks above, plus these 7. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
No API provider serves it, so its downloads are people running it themselves: 22 quantized builds and 92 Spaces.
Top Spaces running it: Tonic/starcoder2 ♥ 80 · KBaba7/Quant ♥ 51 · FallnAI/Quantize-HF-Models ♥ 19 · bhaskartripathi/LLM_Quantization ♥ 16
Weights only, an estimate: leave room for the KV cache. Method
About 9.6 GB at INT4, 19 GB at FP8, 38 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2024-02-20), amber triangles are 9 dated articles (each links to it), grey ticks are repo commits.
56 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Everything else, folded: the full tables, every source.
The architecture utilizes grouped-query and sliding window attention with a context window of 16,384 tokens.[1]
Generated 2026-10-07 15:14 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
Listed in LLMs
| Pipeline tag | text-generation |
|---|---|
| Library | transformers |
| Architectures | Starcoder2ForCausalLM |
| Model type | starcoder2 |
| Dtype | float32 |
| Params by dtype | F32: 15,957,889,024 |
| Size band | L |
| MoE | no |
| Context length | 16,384 (config.json max_position_embeddings) |
| Papers | arXiv:2305.13245, arXiv:2205.14135, arXiv:2004.05150, arXiv:2207.14255, arXiv:2402.19173 |
| License | BigCode OpenRAIL-M (bigcode-openrail-m) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · Starcoder2ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Starcoder2ForCausalLM · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed
vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.
Weights-only memory for starcoder2-15b: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 38 GB · Single high-end GPU (<80 GB) |
|---|---|
| FP8 (1 byte/param) | 19 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 9.6 GB · Single GPU (<24 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
quantized: 22finetune: 21adapter: 13
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| second-state/StarCoder2-15B-GGUF | quantized | GGUF | — | — | bigcode-openrail-m | second-state | 32 | 115.4k |
| lmstudio-community/starcoder2-15b-instruct-v0.1-GGUF | quantized | GGUF | — | — | bigcode-openrail-m | lmstudio-community | 7 | 63k |
| bigcode/starcoder2-15b-instruct-v0.1 | finetune | — | 16.0B | L | bigcode-openrail-m | bigcode | 106 | 56k |
| bartowski/starcoder2-15b-instruct-v0.1-GGUF | quantized | GGUF | — | — | bigcode-openrail-m | bartowski | 9 | 32.6k |
| mradermacher/starcoder2-15b-i1-GGUF | quantized | GGUF | — | — | bigcode-openrail-m | mradermacher | 2 | 22.5k |
| mradermacher/starcoder2-15b-GGUF | quantized | GGUF | — | — | bigcode-openrail-m | mradermacher | 1 | 10.8k |
| akpsahan/starcoder2-15b-Q4_K_M-GGUF | quantized | GGUF | — | — | bigcode-openrail-m | akpsahan | 0 | 9.6k |
| DevQuasar-8/bigcode.starcoder2-15b-GGUF | quantized | GGUF | — | — | — | DevQuasar-8 | 0 | 5.6k |
| LiteLLMs/starcoder2-15b-instruct-v0.1-GGUF | quantized | GGUF | — | — | bigcode-openrail-m | LiteLLMs | 1 | 4.4k |
| tensorblock/starcoder2-15b-GGUF | quantized | GGUF | — | — | bigcode-openrail-m | tensorblock | 0 | 3.8k |
| HuggingFaceH4/starchat2-15b-sft-v0.1 | finetune | — | 16.0B | L | bigcode-openrail-m | HuggingFaceH4 | 5 | 2.4k |
| RedHatAI/starcoder2-15b-quantized.w8a16 | quantized | -quant | 16.0B | L | bigcode-openrail-m | RedHatAI | 0 | 2.1k |
| opencsg/opencsg-starcoder2-15b-v0.1 | finetune | — | 16.0B | L | bigcode-openrail-m | opencsg | 2 | 1.5k |
| teemperor/starcoder2-15b-Q6_K-GGUF | quantized | Q6_K | — | — | bigcode-openrail-m | teemperor | 1 | 1.5k |
| mlx-community/bigcode-starcoder2-15b-4bit | quantized | -4bit | 16.0B | L | bigcode-openrail-m | mlx-community | 0 | 1.3k |
| OpenVINO/starcoder2-15b-int4-ov | finetune | int4 | — | — | bigcode-openrail-m | OpenVINO | 2 | 907 |
| mlx-community/starcoder2-15b-instruct-v0.1-4bit | finetune | -4bit | 16.0B | L | bigcode-openrail-m | mlx-community | 0 | 879 |
| mlx-community/bigcode-starcoder2-15b-8bit | quantized | -8bit | 16.0B | L | bigcode-openrail-m | mlx-community | 0 | 755 |
| mlx-community/starcoder2-15b-instruct-v0.1-8bit | finetune | -8bit | 16.0B | L | bigcode-openrail-m | mlx-community | 0 | 723 |
| mlx-community/starcoder2-15b-instruct-v0.1 | finetune | — | 16.0B | L | bigcode-openrail-m | mlx-community | 0 | 690 |
| osukhoroslov-hw/starcoder2-15b-Q5_K_M-GGUF | quantized | GGUF | — | — | bigcode-openrail-m | osukhoroslov-hw | 0 | 684 |
| mlx-community/bigcode-starcoder2-15b-fp32 | finetune | — | 16.0B | L | bigcode-openrail-m | mlx-community | 0 | 652 |
| OpenVINO/starcoder2-15b-fp16-ov | finetune | -fp16 | — | — | bigcode-openrail-m | OpenVINO | 0 | 642 |
| mlx-community/bigcode-starcoder2-15b-6bit | quantized | — | 16.0B | L | bigcode-openrail-m | mlx-community | 0 | 625 |
| mlx-community/bigcode-starcoder2-15b-fp16 | finetune | -fp16 | 16.0B | L | bigcode-openrail-m | mlx-community | 0 | 588 |
| OpenVINO/starcoder2-15b-int8-ov | finetune | int8 | — | — | bigcode-openrail-m | OpenVINO | 0 | 424 |
| Johnblick187/SmartCoderDense | finetune | — | 4.65B | M | apache-2.0 | Johnblick187 | 0 | 357 |
| YorkieOH10/starcoder2-15b-instruct-v0.1-Q8_0-GGUF | quantized | Q8_0 | — | — | bigcode-openrail-m | YorkieOH10 | 0 | 342 |
| shavera/starcoder2-15b-w4-autoawq-gemm | quantized | awq | 16.0B | L | bigcode-openrail-m | shavera | 0 | 340 |
| neighborwang/ModeliCo-15B | finetune | — | 16.0B | L | apache-2.0 | neighborwang | 2 | 207 |
| li-muyang/starchat2-15b-v0.1 | finetune | — | 16.0B | L | bigcode-openrail-m | li-muyang | 0 | 306 |
| suika556/starcoder2-15b-Q6_K-GGUF | quantized | Q6_K | — | — | bigcode-openrail-m | suika556 | 0 | 295 |
| jacobcarajo/starcoder2-15b-Q5_K_M-GGUF | quantized | GGUF | — | — | bigcode-openrail-m | jacobcarajo | 0 | 274 |
| Ffftdtd5dtft/starcoder2-15b-Q2_K-GGUF | quantized | Q2_K | — | — | bigcode-openrail-m | Ffftdtd5dtft | 0 | 272 |
| KaQyn/peft-lora-starcoder2-15b-unity-copilot | adapter | — | — | — | — | KaQyn | 0 | 154 |
| andstor/bigcode-starcoder2-15b-unit-test-fine-tuning | finetune | — | 16.0B | L | bigcode-openrail-m | andstor | 0 | 151 |
| kevinwsbr/starcoder-vulnfixes | adapter | — | 16.0B | L | bigcode-openrail-m | kevinwsbr | 0 | 126 |
| lainlives/starcoder2-15b-bnb-4bit | quantized | -bnb | 16.1B | L | bigcode-openrail-m | lainlives | 0 | 122 |
| andstor/bigcode-starcoder2-15b-unit-test-lora | adapter | — | — | — | bigcode-openrail-m | andstor | 0 | 103 |
| andstor/bigcode-starcoder2-15b-unit-test-ia3 | adapter | — | — | — | bigcode-openrail-m | andstor | 0 | 97 |
| Date | Commit | SHA |
|---|---|---|
| 2024-06-05 | Update values of `bos_token_id` and `eos_token_id` in generation_config.json (#25) | 46d4474 |
| 2024-03-06 | change widget inference parameters | 995200d |
| 2024-03-04 | add dropout (#13) | acfb201 |
| 2024-03-04 | Update config.json (#9) | 1dab6bb |
| 2024-03-01 | fix paper (#7) | d020785 |
| 2024-02-28 | Fix table of contents (#4) | c78f7e5 |
| 2024-02-28 | Update README.md | e18a21d |
| 2024-02-27 | NVIDIA framework and contribution updates. (#3) | b1bcca5 |
| 2024-02-27 | Update README.md (#2) | 1c2dae0 |
| 2024-02-27 | add model | 56438be |
| 2024-02-27 | fix model name | cb14254 |
| 2024-02-27 | fix typos | eaa8ac9 |
| 2024-02-27 | Update README.md (#1) | 5c5e0d3 |
| 2024-02-20 | initial commit | 44b897e |
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.
arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:42 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:14 UTC · Last verified 2026-10-25
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
