math benchmark
CodeForces Leaderboard
CodeForces is a competitive programming benchmark using problems from the CodeForces platform to evaluate LLM code generation on algorithmic problems rated from 800 to 2400.
Updated Sep 5, 2026
Models17
Model coverage17
MetricNormalized rating
EvidenceB
CodeForces Ranking
Higher normalized rating ranks better on this benchmark.
Sort by
Rank | Model | Normalized rating | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelDE | Normalized rating100.00% | Percentile100.00% | Participants17 | EvidenceC | Evaluated |
| Rank02 | ModelDE | Normalized rating100.00% | Percentile93.75% | Participants17 | EvidenceC | Evaluated |
| Rank03 | ModelDE | Normalized rating93.87% | Percentile87.50% | Participants17 | EvidenceC | Evaluated |
| Rank04 | ModelDE | Normalized rating90.00% | Percentile81.25% | Participants17 | EvidenceC | Evaluated |
| Rank05 | ModelAC | Normalized rating85.10% | Percentile75.00% | Participants17 | EvidenceC | Evaluated |
| Rank06 | ModelAC | Normalized rating82.20% | Percentile68.75% | Participants17 | EvidenceC | Evaluated |
| Rank07 | ModelOP | Normalized rating82.10% | Percentile62.50% | Participants17 | EvidenceC | Evaluated |
| Rank08 | ModelAC | Normalized rating80.70% | Percentile56.25% | Participants17 | EvidenceC | Evaluated |
| Rank09 | ModelDE | Normalized rating79.50% | Percentile50.00% | Participants17 | EvidenceC | Evaluated |
| Rank10 | ModelDE | Normalized rating79.50% | Percentile43.75% | Participants17 | EvidenceC | Evaluated |
| Rank11 | ModelOP | Normalized rating74.33% | Percentile37.50% | Participants17 | EvidenceC | Evaluated |
| Rank12 | ModelDE | Normalized rating70.70% | Percentile31.25% | Participants17 | EvidenceC | Evaluated |
| Rank13 | ModelDE | Normalized rating69.70% | Percentile25.00% | Participants17 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Normalized rating65.90% | Percentile18.75% | Participants17 | EvidenceC | Evaluated |
| Rank15 | ModelDE | Normalized rating64.33% | Percentile12.50% | Participants17 | EvidenceC | Evaluated |
| Rank16 | ModelGO | Normalized rating55.30% | Percentile6.25% | Participants17 | EvidenceC | Evaluated |
| Rank17 | ModelGO | Normalized rating47.63% | Percentile0.00% | Participants17 | EvidenceC | Evaluated |
CodeForces Highlights
The leading models and scores on this benchmark.
CodeForces Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for CodeForces
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is CodeForces?
What CodeForces measures and how its scores work.
A benchmark based on CodeForces problems covering dynamic programming, graph algorithms, data structures, and mathematical problems.
It measures code generation capabilities on algorithmic problems using the Normalized rating metric, reported as a ratio, with standardized evaluation through direct platform submission.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- CodeForces
- Modality
- text
- Primary category
- math
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- overall
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about CodeForces.
Which model scores highest on CodeForces?
DeepSeek-V4-Flash-Max is currently ranked first with 100.00%.
What are the top three models on CodeForces?
The current leaders are DeepSeek-V4-Flash-Max (100.00%), DeepSeek-V4-Pro-Max (100.00%), and DeepSeek-V4-Flash-0423 (93.87%).
Which CodeForces model has the lowest official input price?
DeepSeek-V4-Flash-Max has the lowest matched official input price at $0.14 input / $0.28 output per 1M tokens.
Which models are fastest among CodeForces results?
The fastest matched records are GPT OSS 20B (1,000.00 tok/s via Groq), GPT OSS 120B (500.00 tok/s via Groq), and DeepSeek-V3.2 (97.00 tok/s via DeepInfra).
Does the highest normalized rating result prove overall model quality?
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.