Skip to main content

CloudflareDecision modelActive

Clef-Flash

PopularityTop group about equally popular, 12 models

45out of 100new calculation

based on two data groups, mostly search, range 36 to 55

Summary

Clef-Flash is Cloudflare’s smaller variant with 9 billion parameters. Like Clef, it evaluates supplied state through typed questions and returns Choice, Noul, and Score as probabilities. It does not generate an explanation. Text, JSON, images, and video remain supported inputs.

Qwen3.5-9B and a vision encoder form its base. Cloudflare specifies a 64k-token context. It is available through Workers AI as @cf/cloudflare/clef-flash for $0.09 per 1M input tokens, as well as a local Apache-2.0 checkpoint. The free neuron allowance is shared across the account. Its smaller design targets a faster structured readout, not a chat dialogue.

GraduallyBench Decision

Clef-Flash decides 51 of 60 test cases correctly. The median of the 33 tested decision models is 60.0 %.

Accuracy

85.0 %

51 of 60 test cases

Robustness

86.7 %

26 of 30 reworded variants

Unnecessary reviews

1.9 %

1 of 52 cases without a required review

Response time

599 ms

median, 95 % under 1.3 s, hosted

Accuracy compared

Share of correct decisions in percent, higher is better

20 of 33 models

Jev93.3 %
Kev 9B93.3 %
Clef91.7 %
Clef-Flash85.0 %
Decider 4B85.0 %
d1-2B85.0 %
Lev81.7 %
d1-Omni-2B81.7 %
Kev 4B78.3 %
Jeff 170.0 %
Decider 2B66.7 %
Kev 0.8B50.0 %

As of 11/10/2026. Source: GraduallyBench Decision

Accuracy by task familyThe tick marks the median of all 33 tested models.
Triage75.0 %9 / 12, median 58.3 %
Routing91.7 %11 / 12, median 83.3 %
Constraints83.3 %10 / 12, median 33.3 %
Tables83.3 %10 / 12, median 50.0 %
Evidence91.7 %11 / 12, median 75.0 %

Clef-Flash

Benchmark run
Show test details

Decisions

Cloudflare · German, English

Execution Hosted. Precision not documented.

130 valid outputs from 130 attempted and 130 planned test slots.

Measured duration p50 / p95 599 / 1,315 ms (130 cases). Hosted request including network time. Nearest-rank quantiles.

Automatic decision accuracy
85 %(51 / 60 planned cases)German, English
Automatic decision accuracy, Triage
75 %(9 / 12 planned cases)German, English
Automatic decision accuracy, Routing
91.67 %(11 / 12 planned cases)German, English
Automatic decision accuracy, Selection with constraints
83.33 %(10 / 12 planned cases)German, English
Automatic decision accuracy, Numbers and tables
83.33 %(10 / 12 planned cases)German, English
Automatic decision accuracy, Sufficient evidence
91.67 %(11 / 12 planned cases)German, English
Correct robustness variants
86.67 %(26 / 30 planned cases)German, English
Correct development cases
90 %(36 / 40 planned cases)German, English
Unnecessary reviews
1.92 %(1 / 52 assessable cases without a required review)German, English
Missed required reviews
12.5 %(1 / 8 assessable cases requiring review)German, English

Response times are comparable only within the same environment and form no shared speed ranking. Where used, the test references were not independently reviewed by humans. Test coverage and methods differ by model. Each metric stands on its own; there is no combined overall score across tests. Data compiled on 11/10/2026. How we test · Download all Decision and Audio results as JSON

First published cohorts

GraduallyBench Decision v1

Automatically evaluated against predefined rules.

The pilot covers 60 test cases from 30 scenarios in German and English. Five equally weighted task families form the score from 0 to 100. Failed requests and invalid answers count as unsuccessful decisions. The 30 robustness variants and 40 development cases are reported separately.

The reference answers follow stored rules and calculations. The automatic reference check took place after the model runs. The task wording and translations have not been checked by human reviewers. Results apply to the exact tested model and route.

Protocol decision-choice-automated-v1. No verified provider invoice.

Clef-Flash

Cloudflare · clef-flash · @cf/cloudflare/clef-flash · Tested Nov 5, 2026

Required decisions across the test plan
85 % (51 / 60)
Valid test responses
60 / 60
Correct robustness variants
26 / 30
Correct development cases, outside the test score
36 / 40
Unnecessary escalations
1 / 52
Missed required escalations
1 / 8
Hosted request, median / p95
599 / 1,315 ms (n = 130)
API cost
Not available

Includes network time and client processing.

Results by task family
TaskCorrect / plannedTimeoutsTransport errorsInvalid answers
Triage9 / 12000
Routing11 / 12000
Selection with constraints10 / 12000
Numbers and tables10 / 12000
Sufficient evidence11 / 12000

All benchmark results

Every published result with its source, test setup and comparison models.

Show all 147 published benchmark results

Page 1 of 13

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 0.2.1: BFCL (case exact accuracy) (Show measurement, test conditions, and source)
Source value
98.8
Metric
BFCL (case exact accuracy)
Unit
%
Benchmark version
0.2.1
Category
Decision Index 0.2.1
Direction
Higher is better
Source type
vendor-reported
Evaluator
Cloudflare Clef Flash model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Run=Cloudflare internal evaluation; Benchmark version=0.2.1; Request latency=no
Source Cloudflare Clef Flash model card
Decision Index 0.2.1: ToolRet (nDCG@10) (Show measurement, test conditions, and source)
Source value
66.4
Metric
ToolRet (nDCG@10)
Unit
%
Benchmark version
0.2.1
Category
Decision Index 0.2.1
Direction
Higher is better
Source type
vendor-reported
Evaluator
Cloudflare Clef Flash model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Run=Cloudflare internal evaluation; Benchmark version=0.2.1; Request latency=no
Source Cloudflare Clef Flash model card
Decision Index 0.2.1: API-Bank (accuracy) (Show measurement, test conditions, and source)
Source value
93.1
Metric
API-Bank (accuracy)
Unit
%
Benchmark version
0.2.1
Category
Decision Index 0.2.1
Direction
Higher is better
Source type
vendor-reported
Evaluator
Cloudflare Clef Flash model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Run=Cloudflare internal evaluation; Benchmark version=0.2.1; Request latency=no
Source Cloudflare Clef Flash model card
Decision Index 0.2.1: BANKING77 (macro-F1) (Show measurement, test conditions, and source)
Source value
90.9
Metric
BANKING77 (macro-F1)
Unit
%
Benchmark version
0.2.1
Category
Decision Index 0.2.1
Direction
Higher is better
Source type
vendor-reported
Evaluator
Cloudflare Clef Flash model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Run=Cloudflare internal evaluation; Benchmark version=0.2.1; Request latency=no
Source Cloudflare Clef Flash model card
Decision Index 0.2.1: CLINC150+OOS (macro-F1) (Show measurement, test conditions, and source)
Source value
66.8
Metric
CLINC150+OOS (macro-F1)
Unit
%
Benchmark version
0.2.1
Category
Decision Index 0.2.1
Direction
Higher is better
Source type
vendor-reported
Evaluator
Cloudflare Clef Flash model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Run=Cloudflare internal evaluation; Benchmark version=0.2.1; Request latency=no
Source Cloudflare Clef Flash model card
Decision Index 0.2.1: RouterBench (selected quality) (Show measurement, test conditions, and source)
Source value
79.9
Metric
RouterBench (selected quality)
Unit
%
Benchmark version
0.2.1
Category
Decision Index 0.2.1
Direction
Higher is better
Source type
vendor-reported
Evaluator
Cloudflare Clef Flash model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Run=Cloudflare internal evaluation; Benchmark version=0.2.1; Request latency=no
Source Cloudflare Clef Flash model card
Decision Index 0.2.1: Home appliance simulator (case exact accuracy) (Show measurement, test conditions, and source)
Source value
97.7
Metric
Home appliance simulator (case exact accuracy)
Unit
%
Benchmark version
0.2.1
Category
Decision Index 0.2.1
Direction
Higher is better
Source type
vendor-reported
Evaluator
Cloudflare Clef Flash model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Run=Cloudflare internal evaluation; Benchmark version=0.2.1; Request latency=no
Source Cloudflare Clef Flash model card
Decision Index 0.2.1: SGD/SGD-X (macro-F1) (Show measurement, test conditions, and source)
Source value
34.2
Metric
SGD/SGD-X (macro-F1)
Unit
%
Benchmark version
0.2.1
Category
Decision Index 0.2.1
Direction
Higher is better
Source type
vendor-reported
Evaluator
Cloudflare Clef Flash model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Run=Cloudflare internal evaluation; Benchmark version=0.2.1; Request latency=no
Source Cloudflare Clef Flash model card
Decision Index 0.2.1: ContractNLI (macro-F1) (Show measurement, test conditions, and source)
Source value
84.3
Metric
ContractNLI (macro-F1)
Unit
%
Benchmark version
0.2.1
Category
Decision Index 0.2.1
Direction
Higher is better
Source type
vendor-reported
Evaluator
Cloudflare Clef Flash model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Run=Cloudflare internal evaluation; Benchmark version=0.2.1; Request latency=no
Source Cloudflare Clef Flash model card
Decision Index 0.2.1: ANLI (macro-F1) (Show measurement, test conditions, and source)
Source value
59.1
Metric
ANLI (macro-F1)
Unit
%
Benchmark version
0.2.1
Category
Decision Index 0.2.1
Direction
Higher is better
Source type
vendor-reported
Evaluator
Cloudflare Clef Flash model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Run=Cloudflare internal evaluation; Benchmark version=0.2.1; Request latency=no
Source Cloudflare Clef Flash model card
Decision Index 0.2.1: BPoMP (accuracy) (Show measurement, test conditions, and source)
Source value
95.4
Metric
BPoMP (accuracy)
Unit
%
Benchmark version
0.2.1
Category
Decision Index 0.2.1
Direction
Higher is better
Source type
vendor-reported
Evaluator
Cloudflare Clef Flash model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Run=Cloudflare internal evaluation; Benchmark version=0.2.1; Request latency=no
Source Cloudflare Clef Flash model card
Decision Index 0.2.1: Humicroedit (accuracy) (Show measurement, test conditions, and source)
Source value
75.1
Metric
Humicroedit (accuracy)
Unit
%
Benchmark version
0.2.1
Category
Decision Index 0.2.1
Direction
Higher is better
Source type
vendor-reported
Evaluator
Cloudflare Clef Flash model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Run=Cloudflare internal evaluation; Benchmark version=0.2.1; Request latency=no
Source Cloudflare Clef Flash model card

Page 1 of 13

Popularity compared

How often the model is searched for, used and mentioned, compared with every model of its category.

Popularity index

0 to 100, higher is more popular

20 of 42 models

As of November 2026 · Source: gradually.ai popularity index

Over time

Weekly values from October 18, 2026 to November 20, 2026

Clef-FlashGLiNER2.5-Decide 46Kev 4B 45AutoTrust JEV-27B-VL 45Jev 43

Method changes in the timeline, a jump at these points is no market event:

  1. 1November 1, 2026: Value only with search or usage data, 30-day attention
  2. 2November 1, 2026: Method version 4 (common scale, shrinkage)
  3. 3November 9, 2026: New topics measurement V20 (daily quorum, short forms)

What is gaining popularity right now, across models, vendors, harnesses and AI subscriptions, is collected in All AI trends.

All specifications

Architecture

Model class
Multimodal System One decision model, not a chat model
Source: Model class
Checkpoint
Cloudflare/clef-flash
Source: Checkpoint
Base model
Qwen/Qwen3.5-9B with vision encoder
Source: Base model

Input and output

Input
Text, JSON, images, and video
Source: Input
Context window
64k tokens (provider specification)
Source: Context window

Watermarking and provenance

AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.

AI watermarking
Unknown
no source
Watermarking technique
Not known
no source

Licence and access

Access
Workers AI API and local weights
Source: Access
License
Apache-2.0
Source: License
Shutdown
None announced
no source

API

API model ID
Workers AI: @cf/cloudflare/clef-flash
Source: API model ID

Further details

Output schema
Choice, Noul and Score with probabilities, without free-form answer text
Source: Output schema

Prices

API input price
$0.090 per 1M input tokens (8,182 neurons per 1M)Source
Workers AI billing
10,000 neurons per day free account-wide; Workers Paid is required beyond that quota. $0.011 per 1,000 neurons.Source

Sources and data date

Show all 14 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.

TypeEvidence and data date
Research statusResearch date October 4, 2026. 6 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Unresolved

forecastbench-scale

Note

ForecastBench prints Brier values 13.9 and 10.6 without a stated scale; the raw values remain unchanged and are not divided by 100.

Not published

immutable-model-or-route-revision

Note

The zero-shot IE custom benchmark maps its exact public label to the profile task, but does not publish the loaded weight hash or immutable serving revision.

Not published

sampling-uncertainty

Note

The zero-shot IE small custom pools have no published confidence intervals. This gap is specific to those score cells.

Not published

Immutable Decision Index served checkpoint revision

Note

The current public index gives a model URL and, where available, a served_checkpoint label, but no immutable commit/digest for the exact evaluated weights.

Additional sourceCloudflare announcement (retrieved October 3, 2026) · Cloudflare Clef announcement · Editorial description reviewed October 3, 2026
Additional sourceClef-Flash model card (retrieved October 3, 2026; October 4, 2026) · Clef-Flash model card · Editorial description reviewed October 3, 2026
Additional sourceWorkers AI pricing (retrieved October 3, 2026) · Workers AI pricing · Editorial description reviewed October 3, 2026
Additional sourceZero-shot IE benchmark, pinned public result snapshot (c45a00c25d2731ac7047def97b91421aaff2f611) (retrieved October 17, 2026)
Additional sourceMinh, Tran and Tan, CLEF/CLEF-Flash ultrasound evaluation, arXiv:2610.13675v1, Table 3 (retrieved October 17, 2026; October 23, 2026)
Additional sourceClef eval, Sharad Bapat (retrieved October 17, 2026)
Additional sourceStreamDecisionBench GitHub README (retrieved October 18, 2026)
Additional sourceJev Decision Index public index.json (release-v4) (retrieved October 28, 2026)
Additional sourceOpenRouter: Popularity index (derived), CC BY 4.0
Additional sourceVercel AI Gateway: Popularity index (derived), CC BY 4.0
Additional sourceOllama: cumulative downloads for the popularity index
Additional sourceReplicate: cumulative runs for the popularity index
Additional sourceGoogle Trending Now: limited feed selection for Germany and the US
Additional sourceCreative Commons Attribution 4.0