Benchmark Almanac

GSM-Infinite

DeepSeek-R1 May 2025 8534.9 leads since May 2025
Not in model cards

Models solve generated arithmetic problems across different context lengths and numbers of reasoning steps. The benchmark varies how many irrelevant details appear alongside each problem.

On November 19, 2026, the repository listed DeepSeek-R1 first on its zero-noise table with an average of 8534.88. GPT-o3-mini scored 6931.88.

Released
February 2025
Kind
Reasoning over long documents
Level
School
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Subject
Mathematics
Language
English
In and out
Text → Text
Interaction
Single turn
Items from
Generated
Refresh
One fixed set
Run by
Anyone
Strongest model tested
DeepSeek-R1 May 2025 #122 on the capability index
Last model added
May 2025
Scores refreshed
November 20th, 2026
19
Entries on its own leaderboard, 14 models we track

DeepSeek-R1 leads with 8534.9 among models released by May 2025 DeepSeek-R1 leads with 8534.9 among models released by May 2025

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in May 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 DeepSeek-R1 May 2025, Zero Noise 8534.9
  2. 2 o3-mini Zero Noise 6931.9
  3. 3 o1-mini Zero Noise 4951.1
  4. 4 DeepSeek-V3 Mar 2025, Zero Noise 3606.2
  5. 5 QwQ-32B-Preview Zero Noise 2860.7
  6. 6 Gemini 1.5 Pro Sep 2024, Zero Noise 2841.6
  7. 7 Claude 3.5 Sonnet Zero Noise 2519.5
  8. 8 mistral-large-2411 Zero Noise 2507.6
  9. 9 Qwen2.5-72B Zero Noise 2187.1
  10. 10 GPT-4o Nov 2024, Zero Noise 2096.0
  11. 11 Gemini 1.5 Flash Sep 2024, Zero Noise 1574.3
  12. 12 llama-3.1-70b-instruct Zero Noise 1541.5
  13. 13 MiniMax-Text-01 Zero Noise 1503.2
  14. 14 llama-3.1-405b-instruct Zero Noise 1276.2
  15. 15 GPT-4o mini Zero Noise 1236.5
  16. 16 Claude 3.5 Haiku Zero Noise 911.5
  17. 17 qwen-2.5-7b-instruct Zero Noise 764.1
  18. 18 llama-3.1-8b-instruct Zero Noise 618.3
  19. 19 Jamba 1.5 Large Zero Noise 602.5

Source: Official GSM-Infinite leaderboard

#EntryScoreSettingDateSource
1 DeepSeek-R1 May 2025 8534.9 Zero Noise May 2025 Official leaderboard
2 o3-mini 6931.9 Zero Noise Jan 2025 Official leaderboard
3 o1-mini 4951.1 Zero Noise Sep 2024 Official leaderboard
4 DeepSeek-V3 Mar 2025 3606.2 Zero Noise Mar 2025 Official leaderboard
5 QwQ-32B-Preview 2860.7 Zero Noise Nov 2024 Official leaderboard
6 Gemini 1.5 Pro Sep 2024 2841.6 Zero Noise Sep 2024 Official leaderboard
7 Claude 3.5 Sonnet 2519.5 Zero Noise Jun 2024 Official leaderboard
8 mistral-large-2411 2507.6 Zero Noise Official leaderboard
9 Qwen2.5-72B 2187.1 Zero Noise Sep 2024 Official leaderboard
10 GPT-4o Nov 2024 2096.0 Zero Noise Nov 2024 Official leaderboard
11 Gemini 1.5 Flash Sep 2024 1574.3 Zero Noise Aug 2024 Official leaderboard
12 llama-3.1-70b-instruct 1541.5 Zero Noise Official leaderboard
13 MiniMax-Text-01 1503.2 Zero Noise Jan 2025 Official leaderboard
14 llama-3.1-405b-instruct 1276.2 Zero Noise Official leaderboard
15 GPT-4o mini 1236.5 Zero Noise Jul 2024 Official leaderboard
16 Claude 3.5 Haiku 911.5 Zero Noise Oct 2024 Official leaderboard
17 qwen-2.5-7b-instruct 764.1 Zero Noise Official leaderboard
18 llama-3.1-8b-instruct 618.3 Zero Noise Official leaderboard
19 Jamba 1.5 Large 602.5 Zero Noise Aug 2024 Official leaderboard
#ModelScoreSettingDateSource
1 DeepSeek-R1 May 2025 8534.9 Zero Noise May 2025 Official leaderboard
2 o3-mini 6931.9 Zero Noise Jan 2025 Official leaderboard
3 o1-mini 4951.1 Zero Noise Sep 2024 Official leaderboard
4 DeepSeek-V3 Mar 2025 3606.2 Zero Noise Mar 2025 Official leaderboard
5 QwQ-32B-Preview 2860.7 Zero Noise Nov 2024 Official leaderboard
6 Gemini 1.5 Pro Sep 2024 2841.6 Zero Noise Sep 2024 Official leaderboard
7 Claude 3.5 Sonnet 2519.5 Zero Noise Jun 2024 Official leaderboard
8 Qwen2.5-72B 2187.1 Zero Noise Sep 2024 Official leaderboard
9 GPT-4o Nov 2024 2096.0 Zero Noise Nov 2024 Official leaderboard
10 Gemini 1.5 Flash Sep 2024 1574.3 Zero Noise Aug 2024 Official leaderboard
11 MiniMax-Text-01 1503.2 Zero Noise Jan 2025 Official leaderboard
12 GPT-4o mini 1236.5 Zero Noise Jul 2024 Official leaderboard
13 Claude 3.5 Haiku 911.5 Zero Noise Oct 2024 Official leaderboard
14 Jamba 1.5 Large 602.5 Zero Noise Aug 2024 Official leaderboard

Rankings by part

We read 6 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Symbolic, Medium and Hard. DeepSeek-R1 leads 3 of them, and Gemini 1.5 Pro leads the other 3. Arrows show how far a model moved from its overall rank among the models we track.

Symbolic

14 models we track have a score here.

  1. 1 DeepSeek-R1 May 2025 7280.0
  2. 2 o3-mini 6690.0
  3. 3 o1-mini 5060.0
  4. 4 DeepSeek-V3 Mar 2025 4310.0
  5. 5 QwQ-32B-Preview 3530.0
  6. 6 Gemini 1.5 Pro Sep 2024 2547.0
  7. 7 GPT-4o Nov 2024 ↑2 2379.0
  8. 8 Claude 3.5 Sonnet ↓1 2161.0
  9. 9 Qwen2.5-72B ↓1 2048.0
  10. 10 Gemini 1.5 Flash Sep 2024 1970.0

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
DeepSeek-V3
Long Context Zero Noise
3606.2
Gemini 1.5 Pro
Long Context Zero Noise
2841.6
Qwen2.5-72B
Long Context Zero Noise
2187.1
Gemini 1.5 Flash
Long Context Zero Noise
1574.3
MiniMax-Text-01
Long Context Zero Noise
1503.2
GPT-4o mini
Long Context Zero Noise
1236.5

Source: Official GSM-Infinite leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
DeepSeek-R1 May 2025 Zero Noise 8534.9
o3-mini Zero Noise 6931.9
o1-mini Zero Noise 4951.1
DeepSeek-V3 Mar 2025 Zero Noise 3606.2
QwQ-32B-Preview Zero Noise 2860.7
Gemini 1.5 Pro Sep 2024 Zero Noise 2841.6
Claude 3.5 Sonnet Zero Noise 2519.5
Qwen2.5-72B Zero Noise 2187.1
GPT-4o Nov 2024 Zero Noise 2096.0
Gemini 1.5 Flash Sep 2024 Zero Noise 1574.3
MiniMax-Text-01 Zero Noise 1503.2
GPT-4o mini Zero Noise 1236.5
Gemini 1.5 Pro Sep 2024 Long Context 963.9
Claude 3.5 Haiku Zero Noise 911.5
Qwen2.5-72B Long Context 724.2
Jamba 1.5 Large Zero Noise 602.5
DeepSeek-V3 Mar 2025 Long Context 575.2
Gemini 1.5 Flash Sep 2024 Long Context 509.3
MiniMax-Text-01 Long Context 388.9
GPT-4o mini Long Context 338.1

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
DeepSeek-V3
Long Context Zero Noise
3606.2
Gemini 1.5 Pro
Long Context Zero Noise
2841.6
Qwen2.5-72B
Long Context Zero Noise
2187.1
Gemini 1.5 Flash
Long Context Zero Noise
1574.3
MiniMax-Text-01
Long Context Zero Noise
1503.2
GPT-4o mini
Long Context Zero Noise
1236.5

Source: Official GSM-Infinite leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
DeepSeek-R1 May 2025 Zero Noise 8534.9
o3-mini Zero Noise 6931.9
o1-mini Zero Noise 4951.1
DeepSeek-V3 Mar 2025 Zero Noise 3606.2
QwQ-32B-Preview Zero Noise 2860.7
Gemini 1.5 Pro Sep 2024 Zero Noise 2841.6
Claude 3.5 Sonnet Zero Noise 2519.5
Qwen2.5-72B Zero Noise 2187.1
GPT-4o Nov 2024 Zero Noise 2096.0
Gemini 1.5 Flash Sep 2024 Zero Noise 1574.3
MiniMax-Text-01 Zero Noise 1503.2
GPT-4o mini Zero Noise 1236.5
Gemini 1.5 Pro Sep 2024 Long Context 963.9
Claude 3.5 Haiku Zero Noise 911.5
Qwen2.5-72B Long Context 724.2
Jamba 1.5 Large Zero Noise 602.5
DeepSeek-V3 Mar 2025 Long Context 575.2
Gemini 1.5 Flash Sep 2024 Long Context 509.3
MiniMax-Text-01 Long Context 388.9
GPT-4o mini Long Context 338.1

2 models have held number one on the official leaderboard since its release in February 2025

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI DeepSeek
Own leaderboard Not tested since May 2025
DeepSeek-R1 8534.9

Source: Official GSM-Infinite leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard
0.00 2000.0 4000.0 6000.0 8000.0 10000.0 2024 2025 0.00 2000.0 4000.0 6000.0 8000.0 10000.0 2024 2025

Source: Official GSM-Infinite leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Long context

All 115 in Long context
  • MRCR v2 In use Multi-round coreference: retrieve the n-th of several similar requests hidden in a long synthetic conversation, up to 1M tokens. Muse Spark 1.3 98.5%
  • LongBench v2 In use 503 multiple-choice questions over 8K-2M-word contexts (QA, dialogue, code repos, structured data). Accuracy. Inkling 2 Small 70.6%
  • GraphWalks In use A graph is given as an edge list in long context. The model answers BFS or parent queries, scored by F1. Claude Fable 5 99.96%
  • OpenAI MRCR v2 In use Models read a long synthetic chat that hides 2, 4 or 8 identical writing requests and must reproduce the response to the i-th one. Scored by mean string match ratio per context bin. Gemini 3.7 Flash 89%
  • CL-bench In use Model must learn new knowledge from a supplied long context (rules, procedures, data) and solve tasks. Strict LLM-verified rubrics. GPT-5.4 27.9%
  • AA-LCR 1.1 In use Models answer 100 questions that need reasoning across sets of real documents of about 100K tokens. An LLM equality checker with a fixed system prompt compares each answer with the key. Step 5 Preview 88.3%

Data audit

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 18 months ago, in May 2025, for DeepSeek-R1. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (19 entries).
  • Pass Models scored 14 models we track, plus 5 other entries on its leaderboard.
  • Warning Newest score The newest came 18 months ago, in May 2025, for DeepSeek-R1.
  • Note Frontier models None of the top 10 has a score. The strongest tested is DeepSeek-R1 (#122).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on infiniailab-gsm-infinite-leaderboard.hf.space: 20 rows for models we track and 5 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check