Liquid AIDecision modelOpen weights
d1-2B
Summary
d1-2B is a local decision model with 1.92 billion parameters. It reads text, images, and video and answers typed questions in one forward pass without generating answer tokens. Built on LFM3-VL-2B, it supports choice probabilities, extraction, boxes, and estimation intervals.
The publisher card lists decision and vision suites plus an unweighted average over shared public benchmarks. It also reports millisecond measurements on three hardware platforms. The internal board includes some private tests and is not fully reproducible. These open-weight local tests are not hosted API measurements.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 49 published benchmark results
Page 1 of 5
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
Decision Index public suite (v0.3) (Show measurement, test conditions, and source)
- Source value
- 51.1
- Score
- 51.1
- Metric
- percent
- Unit
- %
- Category
- Decisions
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- LiquidAI/d1-2B official model card at revision cfe5760ff5d5998f91d9ac56bd4b2dbd96996d41
- Status
- active
- Retrieved at
- 2026-10-29
- Methodology
- Vendor-reported result. Liquid AI comparison table; value is from the d1 2B column. Liquid AI decision-model card; one-pass typed decisions, no generated output tokens. Public scores are manufacturer-reported; image suites use at most 1,000 rows per benchmark.
- Context
- Manufacturer-reported; these sources do not provide an independent reproduction of this result.
DecisionBench (Hanno Labs) v1 (Show measurement, test conditions, and source)
- Source value
- 77.1
- Score
- 77.1
- Metric
- percent
- Unit
- %
- Category
- Decisions
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- LiquidAI/d1-2B official model card at revision cfe5760ff5d5998f91d9ac56bd4b2dbd96996d41
- Status
- active
- Retrieved at
- 2026-10-29
- Methodology
- Vendor-reported result. Liquid AI comparison table; value is from the d1 2B column. Liquid AI decision-model card; one-pass typed decisions, no generated output tokens. Public scores are manufacturer-reported; image suites use at most 1,000 rows per benchmark.
- Context
- Manufacturer-reported; these sources do not provide an independent reproduction of this result.
Decision Bench (decisionbench.ai) v4 (Show measurement, test conditions, and source)
- Source value
- 85.2
- Score
- 85.2
- Metric
- percent
- Unit
- %
- Category
- Decisions
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- LiquidAI/d1-2B official model card at revision cfe5760ff5d5998f91d9ac56bd4b2dbd96996d41
- Status
- active
- Retrieved at
- 2026-10-29
- Methodology
- Vendor-reported result. Liquid AI comparison table; value is from the d1 2B column. Liquid AI decision-model card; one-pass typed decisions, no generated output tokens. Public scores are manufacturer-reported; image suites use at most 1,000 rows per benchmark.
- Context
- Manufacturer-reported; these sources do not provide an independent reproduction of this result.
Fast Decisions (Show measurement, test conditions, and source)
- Source value
- 58.8
- Score
- 58.8
- Metric
- percent
- Unit
- %
- Category
- Decisions
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- LiquidAI/d1-2B official model card at revision cfe5760ff5d5998f91d9ac56bd4b2dbd96996d41
- Status
- active
- Retrieved at
- 2026-10-29
- Methodology
- Vendor-reported result. Liquid AI comparison table; value is from the d1 2B column. Liquid AI decision-model card; one-pass typed decisions, no generated output tokens. Public scores are manufacturer-reported; image suites use at most 1,000 rows per benchmark.
- Context
- Manufacturer-reported; these sources do not provide an independent reproduction of this result.
JevBench v1.4.1 (Show measurement, test conditions, and source)
- Source value
- 77.9
- Score
- 77.9
- Metric
- percent
- Unit
- %
- Category
- Decisions
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- LiquidAI/d1-2B official model card at revision cfe5760ff5d5998f91d9ac56bd4b2dbd96996d41
- Status
- active
- Retrieved at
- 2026-10-29
- Methodology
- Vendor-reported result. Liquid AI comparison table; value is from the d1 2B column. Liquid AI decision-model card; one-pass typed decisions, no generated output tokens. Public scores are manufacturer-reported; image suites use at most 1,000 rows per benchmark.
- Context
- Manufacturer-reported; these sources do not provide an independent reproduction of this result.
ImajevBench v2.0-lite (Show measurement, test conditions, and source)
- Source value
- 75.5
- Score
- 75.5
- Metric
- percent
- Unit
- %
- Category
- Decisions
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- LiquidAI/d1-2B official model card at revision cfe5760ff5d5998f91d9ac56bd4b2dbd96996d41
- Status
- active
- Retrieved at
- 2026-10-29
- Methodology
- Vendor-reported result. Liquid AI comparison table; value is from the d1 2B column. Liquid AI decision-model card; one-pass typed decisions, no generated output tokens. Public scores are manufacturer-reported; image suites use at most 1,000 rows per benchmark.
- Context
- Manufacturer-reported; these sources do not provide an independent reproduction of this result.
WorkflowEvals (Show measurement, test conditions, and source)
- Source value
- 81.6
- Score
- 81.6
- Metric
- percent
- Unit
- %
- Category
- Decisions
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- LiquidAI/d1-2B official model card at revision cfe5760ff5d5998f91d9ac56bd4b2dbd96996d41
- Status
- active
- Retrieved at
- 2026-10-29
- Methodology
- Vendor-reported result. Liquid AI comparison table; value is from the d1 2B column. Liquid AI decision-model card; one-pass typed decisions, no generated output tokens. Public scores are manufacturer-reported; image suites use at most 1,000 rows per benchmark.
- Context
- Manufacturer-reported; these sources do not provide an independent reproduction of this result.
Typed Decisions (Show measurement, test conditions, and source)
- Source value
- 61.3
- Score
- 61.3
- Metric
- percent
- Unit
- %
- Category
- Decisions
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- LiquidAI/d1-2B official model card at revision cfe5760ff5d5998f91d9ac56bd4b2dbd96996d41
- Status
- active
- Retrieved at
- 2026-10-29
- Methodology
- Vendor-reported result. Liquid AI comparison table; value is from the d1 2B column. Liquid AI decision-model card; one-pass typed decisions, no generated output tokens. Public scores are manufacturer-reported; image suites use at most 1,000 rows per benchmark.
- Context
- Manufacturer-reported; these sources do not provide an independent reproduction of this result.
Our board (over a hundred public and private benchmarks) (Show measurement, test conditions, and source)
- Source value
- 63.2
- Score
- 63.2
- Metric
- percent
- Unit
- %
- Category
- Decisions
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- LiquidAI/d1-2B official model card at revision cfe5760ff5d5998f91d9ac56bd4b2dbd96996d41
- Status
- active
- Retrieved at
- 2026-10-29
- Methodology
- Vendor-reported result. Liquid AI comparison table; value is from the d1 2B column. Liquid AI decision-model card; one-pass typed decisions, no generated output tokens. Public scores are manufacturer-reported; image suites use at most 1,000 rows per benchmark.
- Context
- Publisher’s internal board evaluation across more than 100 public and private benchmarks. Part of the dataset is private, so the aggregate is not fully independently reproducible.
AI2D (Show measurement, test conditions, and source)
- Source value
- 76.7
- Score
- 76.7
- Metric
- percent
- Unit
- %
- Category
- Vision
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- LiquidAI/d1-2B official model card at revision cfe5760ff5d5998f91d9ac56bd4b2dbd96996d41
- Status
- active
- Retrieved at
- 2026-10-29
- Methodology
- Vendor-reported result. Liquid AI comparison table; value is from the d1 2B column. Liquid AI decision-model card; one-pass typed decisions, no generated output tokens. Public scores are manufacturer-reported; image suites use at most 1,000 rows per benchmark.
- Context
- Manufacturer-reported; these sources do not provide an independent reproduction of this result.
BLINK (Show measurement, test conditions, and source)
- Source value
- 66.2
- Score
- 66.2
- Metric
- percent
- Unit
- %
- Category
- Vision
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- LiquidAI/d1-2B official model card at revision cfe5760ff5d5998f91d9ac56bd4b2dbd96996d41
- Status
- active
- Retrieved at
- 2026-10-29
- Methodology
- Vendor-reported result. Liquid AI comparison table; value is from the d1 2B column. Liquid AI decision-model card; one-pass typed decisions, no generated output tokens. Public scores are manufacturer-reported; image suites use at most 1,000 rows per benchmark.
- Context
- Manufacturer-reported; these sources do not provide an independent reproduction of this result.
CountQA (Show measurement, test conditions, and source)
- Source value
- 51.4
- Score
- 51.4
- Metric
- percent
- Unit
- %
- Category
- Vision
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- LiquidAI/d1-2B official model card at revision cfe5760ff5d5998f91d9ac56bd4b2dbd96996d41
- Status
- active
- Retrieved at
- 2026-10-29
- Methodology
- Vendor-reported result. Liquid AI comparison table; value is from the d1 2B column. Liquid AI decision-model card; one-pass typed decisions, no generated output tokens. Public scores are manufacturer-reported; image suites use at most 1,000 rows per benchmark.
- Context
- Manufacturer-reported; these sources do not provide an independent reproduction of this result.
Page 1 of 5
All specifications
Architecture
- Architecture
- Dense LFM3-VL architecture with 24 layers: 13 gated short convolutions, 5 sliding-window attention layers, and 6 global attention layers.
- Source: Architecture
Input and output
- Output modalities
- Typed, calibrated answers in one forward pass, with no generated output tokens
- Source: Output modalities
Licence and access
- Access
- Open weights for local deployment. Publisher documentation lists d1-2B as a local open-weight model.
- Source: Access
Further details
- Backbone
- Built on LFM3-VL-2B. Results for the underlying vision-language model are not d1-2B benchmark results.
- Source: Backbone
- Answer types
- Supports classification and typed answers, including choice probabilities, text, numbers, dates, boxes, and intervals.
- Source: Answer types
- Deployment
- Official decision-model documentation distinguishes local open-weight d1 models from hosted d1.1 and d1-ultra API models.
- Source: Deployment
Sources and data date
Show all 9 sources
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.
Without vendor claims and historical or disputed results; those stay labelled in the single values.
| Type | Evidence and data date |
|---|---|
| Research status | Research date October 29, 2026. 61 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Not found in the inspected sources | Technical paper or report Note The model card cites blog posts but does not link a separate technical report or research paper. |
| Not found in the inspected sources | Independent evaluation Note The reviewed manufacturer sources do not link an independent reproduction of these d1 results. |
| Unresolved | Additional latency methodology Note The card specifies warm-up, bfloat16, 20 measurements and a best-of-three selection for H100 and M5 Pro. The aggregation for MI325X is not separately described. |
| Not found in the inspected sources | Vals rows in captured payloads Note Nine complete Vals payloads dated 28 October were checked; no exact model ID or unambiguous model label. Limited to these saved benchmarks. |
| Unresolved | Uncaptured Arena Code/Speech leaderboards Note Text, Agent including Code/Chat/Work, and twelve WebDev views were captured. A separate general Code and Arena Speech leaderboard were not saved as complete tables. |
| Unresolved | Artificial Analysis catalog reconciliation Note The full current AA model and benchmark catalog was not re-reconciled for this profile in this sidecar. This sidecar makes no AA absence claim; native AA imports remain authoritative. |
| Additional source | LiquidAI/d1-2B model card at pinned revision cfe5760ff5d5998f91d9ac56bd4b2dbd96996d41 (retrieved October 29, 2026) · d1-2B model card · Editorial description reviewed October 29, 2026 |
| Additional source | Liquid AI LFM3 release announcement (retrieved October 29, 2026) · Liquid AI LFM3 release announcement · Editorial description reviewed October 29, 2026 |
| Additional source | Liquid AI model documentation (retrieved October 29, 2026) · Liquid AI decision models documentation · Editorial description reviewed October 29, 2026 |
| Additional source | Liquid AI LFM3 Hugging Face collection (retrieved October 29, 2026) |
| Additional source | Hugging Face model metadata (retrieved October 29, 2026) |
| Additional source | Model metadata (retrieved October 29, 2026) · Liquid AI decision models documentation · Editorial description reviewed October 29, 2026 |
| Additional source | Model metadata (retrieved October 29, 2026)123 |