Skip to main content
Radar

Every change to the models you run, with its source and its date. Releases, price changes, retirements, API changes, and incidents.Every change to the models you run, with its source.

Follow model changes
BenchLM

Grade School Math 8K (GSM8K)

A grade-school mathematical reasoning benchmark reported in DeepSeek-V4 base-model evaluations.

Data verified 33 confirmed releases in the last 30 daysSee provider release alerts

Benchmark score on GSM8K — September 15, 2026

We mirror the published score view for GSM8K. Celeris-1 leads the public snapshot at 93.7%, followed by DeepSeek V4 Pro Base (92.6%) and DeepSeek V4 Flash Base (90.8%). We do not use these results to rank models overall.

1Proprietary

Celeris-1

Celeris

93.7%
Context 128K
2Open weights

DeepSeek V4 Pro Base

DeepSeek

92.6%
Context 1M
3Open weights

DeepSeek V4 Flash Base

DeepSeek

90.8%
Context 1M

4 modelsMathCurrentDisplay onlyUpdated September 15, 2026

Benchmark score table (4 models)

Score
1
Celeris-1Celeris · Closed
93.7%
2
DeepSeek V4 Pro BaseDeepSeek · Open weight
92.6%
3
DeepSeek V4 Flash BaseDeepSeek · Open weight
90.8%
4
Soofi S 30B-A3BSoofi Project · Open weight
86.1%

The published GSM8K snapshot places Celeris-1 first at 93.7%. The third row is 2.9 points behind. The broader top-10 range is 7.6 points, so many of the published results sit in a relatively narrow band.

4 models have been evaluated on GSM8K. The benchmark falls in the Math category. This category carries a 5% weight in BenchLM.ai's overall scoring system. GSM8K is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About GSM8K

Year

2026

Tasks

Grade-school math word problems

Format

Exact match

Difficulty

Grade-school math

BenchLM stores GSM8K as a display-only provider-table row when exact values are published in DeepSeek-V4 evaluations.

DeepSeek-V4 Technical Report

BenchLM freshness & provenance

Version

GSM8K 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

FAQ

What does GSM8K measure?

A grade-school mathematical reasoning benchmark reported in DeepSeek-V4 base-model evaluations.

Which model scores highest on GSM8K?

Celeris-1 by Celeris currently leads with a score of 93.7% on GSM8K.

How many models are evaluated on GSM8K?

4 AI models have been evaluated on GSM8K on BenchLM.

Last updated: September 15, 2026 · BenchLM version GSM8K 2026

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 2,000+ readers.

One email each week. Unsubscribe anytime.