DeepSeek

DeepSeek V3.2

Announced Dec 1, 2025

Updated reasoning modes

What could it do?

DeepSeek upgraded its chat and reasoner endpoints to V3.2.

What changed?

Both aliases used the new model.

WHY IT MATTERED

Updated capabilities reached existing API users.

Where it fell short

Speciale was a separate temporary endpoint.

Sources checked Oct 7, 2026

Release facts were checked against the sources below. Performance claims belong to the developers; we have not independently tested these models.

Developer announcement or release log

Benchmark results

EPOCH AI CAPABILITY ESTIMATE

146.3index points

Tested variant: DeepSeek-V3.2Source interval: 144.5 to 147.4Variant date in source: Dec 1, 2025

A benchmark estimate, not a percentage or capability multiplier. Reasoning settings are not specified in this source table. Historical estimates can change in later snapshots.

Epoch AI methodology ↗Download the source snapshotChecked Oct 7, 2026 · CC BY 4.0

PUBLISHED BENCHMARK RESULT

1423rating points

Tested: deepseek-v3.2-thinkingText Arena Overall · October 8, 2026Reported interval: 1419 to 142742,076 votes

One explicitly named variant per release. Scores come from the same Overall snapshot; preliminary entries and reported intervals are preserved. These are current ratings of earlier variants, not their launch day ratings.

Source: Text Arena ↗Download selected resultsChecked Oct 8, 2026

PUBLISHED BENCHMARK RESULT

70.0%% resolved

Tested: DeepSeek V3.2 (high)500 tasks · mini-SWE-agent v2.0.0Run: Feb 17, 2026 · Effort: highPublished submission; not marked as checked by the benchmark team

Only mini-SWE-agent v2.0.0 submissions, one run per model. Reasoning effort varies and is labelled. These are published submissions, not a rerun by Leapscope. Latest included release: February 5, 2026; latest run: February 26, 2026.

Source: SWE bench Verified ↗Download selected resultsChecked Oct 8, 2026

Understand the benchmarks Find this release on the chart