The story so far
Weights, every dated article, and repo commits on one rail.
17 repo commits since the weights landed
Teal is the weights (2026-09-21), grey ticks are repo commits.
CLM-v0.1-8B
Contrastive-LM Apache 2.0 out 34 days ago
A query and passages in, relevance scores out.
Verified 2026-10-26 07:27 UTC Rank #508 in Decision models · score 55.5 method v1 Sources 3 Downloads 30d 4.9k · #546 of 619 tracked
Do people use it?4.9kdownloads, 30 days#546 of 619 we track · below the median+5 likes, downloads −1% this week785 likesHugging Face2026-10-26
Where it's used
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: 9 quantized builds and 8 Spaces.
- As a service0no API provider listed
- On own hardware9quantized builds
- Built on99 fine-tunes · 0 adapters · 0 merges
- In apps8Hugging Face Spaces
Top Spaces running it: multimodalart/jev-decision-index ♥ 802
What people built on it
18 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
18 derivatives on the Hub, most of them quantized
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Reference
Everything else, folded: the full tables, every source.
Benchmarks
2 rows
Benchmarks 2 rows
measured third-party leaderboard rowauthor-reported number from the model card or eval files (2)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.
| Benchmark | Value | Setting / notes | Reported by | Date |
|---|---|---|---|---|
| author-reported DeepSWE · Write code with agents | 81.6% percent | Fine-tuned as a verifier | model author | 2026-10-07 |
| author-reported Terminal-Bench · Write code with agents | 87.6% percent | Fine-tuned as a verifier | model author | 2026-10-07 |
What the card says
1 cited sentence
Summary generated, cited
The architecture uses two projection heads on top of a frozen encoder to connect states and actions.[1]
Generated 2026-10-07 15:17 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
Use-case evidence
3 use cases
What it is good at 3 use cases
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.
Facts
13 fields
Listed in Decision models · Embeddings
Facts
- License: Apache 2.0 (friction: none). Permissive; patent grant. (license file, observed 2026-10-26)
- Created 2026-09-21, last modified 2026-09-24. (HF API, observed 2026-10-26)
| Pipeline tag | text-ranking |
|---|---|
| Library | contrastive-lm |
| Model type | clm |
| MoE | no |
| Languages | en |
| License | Apache 2.0 (apache-2.0) |
| Hub snapshot | 2026-10-26 |
Servability
SGLang
Servability
SGLang: not in registry · 60d941293e3af27a12c6eb64e2c3e2e1a17f100f
SGLang registry (60d941293e3af27a12c6eb64e2c3e2e1a17f100f) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.
Decision facts
catalog
Decision-model catalog curated
| How it was made | frozen Qwen3-8B + contrastive state/action projection heads (new method) |
|---|---|
| Lineage bucket | F - fine-tune of an open model (public/synthetic data) |
| Lineage evidence | model card + HF base_model |
| Base model | Qwen/Qwen3-8B-Base |
| DI 0.2.1 score | 7.40 (independent (DI 0.2.1 board)) |
| ECE (board) | 0.3227 |
| Median latency | 46.8 ms (DI board, 1x RTX PRO 6000, single process) |
| Hosted price (input) | self-hosted/M |
| Source | https://huggingface.co/Contrastive-LM/CLM-v0.1-8B |
All variants
18 variants
Lineage
Base model: Qwen/Qwen3-8B
quantized: 9finetune: 9
Top variants (18 shown, ranked by downloads)
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| czl/CLM-v0.1-8B-GGUF | quantized | GGUF | — | — | apache-2.0 | czl | 13 | 5.7k |
| czl/CLM-v0.1-8B-MLX-8bit | quantized | MLX | 8.19B | M | apache-2.0 | czl | 0 | 247 |
| czl/CLM-v0.1-8B-MLX | finetune | MLX | 8.19B | M | apache-2.0 | czl | 0 | 195 |
| czl/CLM-v0.1-8B-MLX-4bit | quantized | MLX | 8.19B | M | apache-2.0 | czl | 0 | 143 |
| czl/CLM-v0.1-8B-MLX-6bit | quantized | MLX | 8.19B | M | apache-2.0 | czl | 0 | 130 |
| abe238/clm-tune-mlx | finetune | mlx | — | — | apache-2.0 | abe238 | 2 | 0 |
| EvoAwaken-Workshop/CLM-v0.1-8B-gguf | quantized | gguf | — | — | apache-2.0 | EvoAwaken-Workshop | 0 | 84 |
| czl/CLM-v0.1-8B-MLX-8bit-outq2 | quantized | MLX | 8.19B | M | apache-2.0 | czl | 0 | 75 |
| czl/CLM-v0.1-8B-MLX-6bit-outq2 | quantized | MLX | 8.19B | M | apache-2.0 | czl | 0 | 61 |
| czl/CLM-v0.1-8B-MLX-4bit-outq2 | quantized | MLX | 8.19B | M | apache-2.0 | czl | 0 | 61 |
| mudler/CLM-v0.1-8B-vllm-cpp | finetune | — | 8.19B | M | apache-2.0 | mudler | 0 | 41 |
| ahmedandaloes/CLM-v0.1-8B-heads-CoreAI | finetune | — | — | — | apache-2.0 | ahmedandaloes | 0 | 21 |
| bandlagom/clm8b-treet-kd-verdict-demo | finetune | — | — | — | odbl | bandlagom | 0 | 0 |
| AbeHou/harvey-clm-head | finetune | — | — | — | — | AbeHou | 0 | 0 |
| ollaya-dev/clm | quantized | — | — | — | apache-2.0 | ollaya-dev | 0 | 0 |
| tt-hous/clm-v0.1-8b | finetune | — | — | — | apache-2.0 | tt-hous | 0 | 0 |
| harvinderbhullar/CLM-vision-fracatlas | finetune | — | — | — | other | harvinderbhullar | 0 | 0 |
| dotimothy/sima-clm | finetune | — | — | — | apache-2.0 | dotimothy | 0 | 0 |
Repository changelog
17 commits
Repository changelog 17 commits
| Date | Commit | SHA |
|---|---|---|
| 2026-09-24 | Keep --max-model-len 2048 so vLLM fits on a 24 GB GPU | e939398 |
| 2026-09-24 | Simplify fine-tuning command; fetch heldout_tasks.json | 0dea164 |
| 2026-09-24 | Set library_name: contrastive-lm (download stats) | c711783 |
| 2026-09-24 | Add config.json (enables Hub download stats) | c0e58e8 |
| 2026-09-24 | Install via pip install contrastive-lm; add LICENSE | 8597e05 |
| 2026-09-24 | Add license section to model card | 87655cb |
| 2026-09-24 | Add Apache 2.0 license | f1cbaaa |
| 2026-09-24 | Move playground screenshot after fine-tuning | c0fc8ca |
| 2026-09-24 | Add playground screenshot to model card | e772b64 |
| 2026-09-24 | Add playground screenshot | 65eed4a |
| 2026-09-24 | Remove text-only limitation | 44b1cf0 |
| 2026-09-24 | Rename caching bullet | ff7012a |
| 2026-09-24 | Update model card intro | 4add277 |
| 2026-09-24 | Model card: drop mid-training from summary | 0da7540 |
| 2026-09-24 | Add model card | e3eca75 |
| 2026-09-21 | Upload CLM_v0.1-8B.pt | a6cabbc |
| 2026-09-21 | initial commit | f6972c8 |
Similar models
3 models
Similar models
- bge-m3 more providers, permissive license
- bge-large-zh-v1.5 more providers, permissive license
- codebert-base more providers
Sources
3 sources
Sources
Hub snapshot: 2026-10-26 · HF API: 2026-10-26 07:25 UTC · Providers: 2026-10-26 07:26 UTC · Articles: 2026-10-26 07:27 UTC · Summary: 2026-10-07 15:17 UTC · Last verified 2026-10-26 07:27 UTC
Follow CLM-v0.1-8B
Watch this model
- /feeds/models/Contrastive-LM--CLM-v0.1-8B.xml(RSS 2.0) · every change to CLM-v0.1-8B
- /feeds/families/Contrastive-LM--CLM-v0.1.xml(RSS 2.0) · every change across the CLM-v0.1 family
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.