The story so far
Weights, every dated article, and repo commits on one rail.
3 repo commits since the weights landed
Teal is the weights (2026-10-01), grey ticks are repo commits.
Models › pplx-decider-v1 family
pplx-decider-v1-27b
perplexity-ai Apache 2.0 out 24 days ago
26.1B parameters, text in, labels out, 256K tokens of context.
Verified 2026-10-25 Rank #0 in Decision models · score 0 method v1 Sources 3 Downloads 30d 3.5k · #564 of 619 tracked
Do people use it?3.5kdownloads, 30 days#564 of 619 we track · below the median+7 likes, downloads +17% this week132 likesHugging Face2026-10-24
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: 1 quantized builds and 2 Spaces.
Will it fit?
Weights only, an estimate: leave room for the KV cache. Method
Fits a 16 GB card at INT4
About 16 GB at INT4, 31 GB at FP8, 63 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
What people built on it
3 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
3 derivatives on the Hub, led by 1 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
12 rows
Benchmarks 12 rows
measured third-party leaderboard rowauthor-reported number from the model card or eval files (12)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 BBH · Reason through hard problems | 82.80% percent | — | model author | 2026-10-14 |
| author-reported Belebele · Work across languages | 94.00% percent | — | model author | 2026-10-14 |
| author-reported Circa | 89.20% percent | — | model author | 2026-10-14 |
| author-reported ContractNLI | 80.78% percent | — | model author | 2026-10-14 |
| author-reported FinancialPhraseBank | 84.18% percent | — | model author | 2026-10-14 |
| author-reported JevBench public hard | 70.30% percent | — | model author | 2026-10-14 |
| author-reported JudgeBench | 78.29% percent | — | model author | 2026-10-14 |
| author-reported Overall | 85.71% percent | — | model author | 2026-10-14 |
| author-reported RAGTruth | 88.80% percent | — | model author | 2026-10-14 |
| author-reported TabFact | 90.60% percent | — | model author | 2026-10-14 |
| author-reported TruthfulQA · Chat and follow instructions | 85.40% percent | — | model author | 2026-10-14 |
| author-reported Winogrande · Chat and follow instructions | 83.30% percent | — | model author | 2026-10-14 |
Use-case evidence
7 use cases
What it is good at 7 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.
- Chat and follow instructionsauthor-reportedAuthor reports WinoGrande 83.30%Author reports TruthfulQA binary 85.40%chat template present
- Reason through hard problemsauthor-reportedAuthor reports BBH 82.80%chat template: <think>
- Work across languagesauthor-reportedAuthor reports Belebele 94.00%
- Classify and make typed decisionsstructuraldecision-model estimate p=1.00tag: text-classificationpipeline: text-classification+1 more
- Understand images and videostructuraltag: multimodalconfig: vision_config
- Call tools and APIsstructuralchat template: <tool_call>
- Read long documentsstructuralconfig: 262,144 context
Facts
20 fields
Listed in Decision models
Facts
- pplx-decider-v1-27b has 26.1B parameters. (safetensors metadata, observed 2026-10-24)
- Context length is 262,144 tokens (256K). (config.json max_position_embeddings, observed 2026-10-24)
- License: Apache 2.0 (friction: none). Permissive; patent grant. (license file, observed 2026-10-24)
- Created 2026-10-01, last modified 2026-10-12. (HF API, observed 2026-10-24)
| Pipeline tag | text-classification |
|---|---|
| Library | pytorch |
| Architectures | Qwen3_5Model |
| Model type | qwen3_5 |
| Dtype | bfloat16 |
| Params by dtype | BF16: 26,085,330,160 |
| Size band | L |
| MoE | no |
| Context length | 262,144 (config.json max_position_embeddings) |
| License | Apache 2.0 (apache-2.0) |
| Hub snapshot | 2026-10-25 |
Servability
vLLM · SGLang
Servability
vLLM: not in registry · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen3_5Model · 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.
Hardware fit
3 estimates
Hardware fit ESTIMATE
Weights-only memory for pplx-decider-v1-27b: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 63 GB · Single high-end GPU (<80 GB) |
|---|---|
| FP8 (1 byte/param) | 31 GB · Single high-end GPU (<80 GB) |
| INT4 (0.5 bytes/param) | 16 GB · Single GPU (<24 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
Decision facts
catalog
Decision-model catalog curated
| How it was made | re-hosting / rebrand of AutoJev-27B |
|---|---|
| Lineage bucket | R - re-upload / rebrand of another decision model |
| Lineage evidence | third-party SHA-256 comparison (modelsystem.one) + NOTICE file; HF base_model Qwen/Qwen3.8-27B |
| Base model | AutoJev-27B (byte-identical files) |
| DI 0.2.1 score | 56.40 (independent (board row for the same weights)) |
| ECE (board) | 0.0180 |
| Median latency | 101 ms (DI board) |
| Hosted price (input) | $0.04/M |
| Notes | Perplexity launch post does not mention AutoJev |
| Source | https://huggingface.co/perplexity-ai/pplx-decider-v1-27b |
All variants
3 variants
Lineage
Base model: Qwen/Qwen3.8-27B
quantized: 1finetune: 1merge: 1
Top variants (3 shown, ranked by downloads)
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| ramgpt/pplx-decider-v1-27b-EXL3 | quantized | EXL3 | 7.67B | M | apache-2.0 | ramgpt | 3 | 132 |
| ArthurZ/pplx-decider-v1-27b | finetune | — | 1.31M | S | apache-2.0 | ArthurZ | 0 | 41 |
| AveLabs/Qwen3.8-27B-Perplexing-Pegasus | merge | — | 27.8B | L | apache-2.0 | AveLabs | 0 | 36 |
Repository changelog
3 commits
Similar models
3 models
Similar models
- rune-26b-a4b-GGUF · 25.8B same size band (L), same task (text-classification), permissive license
- JEV-27B · 26.9B same size band (L), same task (text-classification), permissive license
- blink-27b · 26.9B same size band (L), same task (text-classification)
Sources
3 sources
Sources
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:43 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:18 UTC · Last verified 2026-10-25
Follow pplx-decider-v1-27b
Watch this model
- /feeds/models/perplexity-ai--pplx-decider-v1-27b.xml(RSS 2.0) · every change to pplx-decider-v1-27b
- /feeds/families/perplexity-ai--pplx-decider-v1.xml(RSS 2.0) · every change across the pplx-decider-v1 family
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.