Provider
DeepSeek
Base model family
Model intelligence profile
DeepSeek V3.2 benchmark results and model details
This family profile brings together 3 published benchmark records for DeepSeek V3.2. Every result keeps its original benchmark, configuration, scale, and source; unrelated scores are never averaged.
Published benchmark evidence available · Business Skills V3 not yet authorized
Release date
Not yet verified
Exact-family source only
Published benchmarks
3
Native scales kept separate
Artificial Analysis
No exact match
As of 2026-07-16
Benchmark-level evidence
Published DeepSeek V3.2 benchmark results
These are individual benchmarks, not collection rollups. Scores remain on their original scales, and agent or harness results stay labelled as system configurations.
Spring Prompt benchmark
PredictTheWeek pilot
10.0%
Mean prediction score
Can LLMs anticipate next week’s Guardian agenda from last week’s coverage?
This benchmark is not on a weekly live cadence yet. The table and charts reflect a single multi-model comparison on one forecast window—a pilot snapshot, not an updating leaderboard. The evaluation setup (clustering, prompts, automated judge, and scoring checks) is still being refined; reported scores and details may change as we improve the pipeline.
1 published configuration
DeepSeek V3.2
10.0%
- Accuracy (partial support or better)
- 13.0%
- Outcome-line coverage
- 3.1%
- Pilot composite
- 6.6%
Spring Prompt benchmark
ROASBench
19.77
Average ROASBench score
A 12-month performance-marketing simulation scored on business outcomes, planning, behavior, and persona fit.
1 published configuration
DeepSeek: DeepSeek V3.2
19.77
- Average contribution profit
- $-126,535
- Average ROAS
- 1.20×
- Completed runs
- 3
- Score variability
- ±3.94
Source: Spring Prompt
Recorded 25 Jul 2026
Published ROASBench cache · 12-month simulation
Source record →
source-native benchmark
Repository Issue Resolution
71.6%
OpenHands SWE-Bench resolved
Direct OpenHands SWE-Bench outcomes for resolving real repository issues with a pinned OpenHands agent and language-model configuration.
Scores are official OpenHands SWE-Bench resolved percentages for exact OpenHands and model configurations, not bare-model scores or SpringPrompt predictions. Twenty-nine exact task sidecars support marginal 95% bootstrap score intervals; five rows explicitly have no interval. No full-cohort rank confidence is claimed.
1 published configuration
OpenHands v1.8.3 + DeepSeek-V3.2-Reasoner
71.6%
- 95% task-bootstrap interval
- Unavailable
- Task evidence
- not-eligible-for-task-bootstrap
Source: OpenHands Index
Recorded 30 Jun 2026
OpenHands · SWE-Bench 2026.06.30-3015ac6
Source record →
Independent operational reference
Artificial Analysis details
No exact, reviewed Artificial Analysis operational record is attached to this family. We do not inherit price, release, or throughput values from a nearby model name.
Early-user signal
Initial community opinions
No editor-reviewed community-opinions paragraph is active for this model. An active paragraph requires at least two retained Reddit discussions from the minimum 14-day launch window—implemented exactly as release date through day 14, end-exclusive—plus paraphrase-only review and a sealed publication artifact. Community reports stay separate from every benchmark score and model comparison.
Independent early-test signal
Early technical field tests
An editorial paraphrase of 2 launch-window field tests from 2 independent authors on X, including 2 reports with a described method or inspectable artifact. The window runs from 1 Dec 2025 up to 15 Dec 2025, and exact model identity was editor reviewed.
Early DeepSeek V3.2 evidence covered two narrow reasoning checks. A procedurally generated, out-of-distribution logic benchmark produced a favorable result within its tested open-model set, while a separate user shared the output from one International Mathematical Olympiad challenge. Both are useful snapshots, but neither demonstrates tool-use reliability or consistent production behavior.
These reports are selectively surfaced and are not a representative sample. They never affect benchmark scores, rankings, winners, or comparison outcomes.
Business Skills V3 · proposed
How Spring Prompt plans to test DeepSeek V3.2
The setup below is proposed and may change until preflight and execution approval are complete. Existing benchmark evidence above does not authorize or stand in for a Business Skills V3 result.
- Proposed reasoning
- High reasoning
- Provider revision
- Exact provider revision will be resolved and frozen only after preflight and execution approval
- Tools and service
- No tools · Provider-default service tier
- Planning configuration
- deepseek-v3.2-high
- Generation controls
- Provider-managed reasoning; no temperature override proposed
Future first-party coverage
15 proposed Business Skills V3 task areas
These links describe evaluation contracts, not published DeepSeek V3.2 results.
Professional Knowledge Work & StrategyBusiness Workflow AutomationCustomer Support & Service ResolutionSales & MarketingLegal Document WorkFinance, Accounting & SpreadsheetsData Analysis & SQLResearch & Evidence SynthesisBusiness Writing & EmailDocument Comprehension & ExtractionSummarization & Grounded KnowledgeSoftware Engineering & CodingFrontend, UI & Web AppsTool Calling & Structured OutputPresentations, Charts & Visual Communication
Model comparisons
Compare DeepSeek V3.2 side by side
Editorially reviewed comparisons appear first. Every page matches only benchmark records with the same reviewed protocol key and does not manufacture an overall winner.
Publication safeguards
What must pass before a V3 result appears
- 1First-party task-local comparisons and eligible external evidence must both be present.
- 2Model and provider configuration identity must match the reviewed release exactly.
- 3Coverage, reliability, judge diagnostics, and sealed stability checks must pass.
- 4Uncertainty and missing evidence remain visible when results are published.
Stable family URL
Evidence can grow without changing the page
New reviewed benchmark snapshots, operational facts, community themes, and early field-test syntheses can be added here while the canonical model-family identity remains fixed.
More from DeepSeek