Spatial Pathfinding Reasoning Challenge
Models trace paths across grids while satisfying rules about cells and edges. The scorer checks each coordinate sequence against valid solutions.
Spatial-Gym later reused the puzzles in a stepwise environment. The original benchmark scores a completed path.
- Made by
- Lars Benedikt Kaesberg and colleagues
- Released
- May 2025
- Kind
- Puzzles and chess
- Level
- School
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- Strongest model tested
- GPT-6.1 Sol #6 on the capability index
- Last model added
- September 2026
- Scores refreshed
- November 20th, 2026
GPT-6.1 Sol leads with 83.6% among models released by September 2026 GPT-6.1 Sol leads with 83.6% among models released by September 2026
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. 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 September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6.1 Terra
- and 15 more
- 83.6 GPT-6.1 Sol
- 41.8
DeepSeek V4.1 Flash
- 38.0 Qwen 3.5 397B (AWQ)
- 21.6 gpt-oss-120b
- 16.6
Gemma 4 31B IT
- 16.2 OLMo 3.1 32B
- 15.8 o4-mini
- 12.6
Qwen3-14B
- 11.4 Nemotron 49B
- 10.0
Qwen3-8B
- 8.40 GLM-4.6V
- 8.20 o3-mini
- 7.80
Qwen3-4B
- 5.80
QwQ-32B
- 5.20
Qwen3-32B
- 4.00
DeepSeek-R1-Distill-Llama-70B
- 3.80
Llama 4 Scout
- 3.60
DeepSeek-R1-Distill-Qwen-32B
- 1.60 GPT-4.1
- 1.60 Magistral Small
- 1.20
Gemma 3 27B
- 1.20
Mistral Small 3.2
- 0.40
Qwen2.5-72B
- 0.40
Qwen3-0.6B
- 1 GPT-6.1 Sol single-shot, medium 83.6%
- 2
DeepSeek V4.1 Flash single-shot, medium 41.8%
- 3 Qwen 3.5 397B (AWQ) single-shot 38%
- 4 gpt-oss-120b single-shot 21.6%
- 5
Gemma 4 31B IT single-shot 16.6%
- 6 OLMo 3.1 32B single-shot 16.2%
- 7 o4-mini single-shot 15.8%
- 8
Qwen3-14B single-shot 12.6%
- 9 Nemotron 49B single-shot 11.4%
- 10
Qwen3-8B single-shot 10%
- 11 GLM-4.6V single-shot 8.40%
- 12 o3-mini single-shot 8.20%
- 13
Qwen3-4B single-shot 7.80%
- 14
QwQ-32B single-shot 5.80%
- 15
Qwen3-32B single-shot 5.20%
- 16
DeepSeek-R1-Distill-Llama-70B single-shot 4.00%
- 17
Llama 4 Scout single-shot 3.80%
- 18
DeepSeek-R1-Distill-Qwen-32B single-shot 3.60%
- 19 GPT-4.1 single-shot 1.60%
- 20 Magistral Small single-shot 1.60%
- 21
Gemma 3 27B single-shot 1.20%
- 22
Mistral Small 3.2 single-shot 1.20%
- 23
Qwen2.5-72B single-shot 0.40%
- 24
Qwen3-0.6B single-shot 0.40%
Source: Official Spatial Pathfinding Reasoning Challenge leaderboard
- 83.6 GPT-6.1 Sol
- 41.8
DeepSeek V4.1 Flash
- 21.6 gpt-oss-120b
- 16.6
Gemma 4 31B IT
- 15.8 o4-mini
- 12.6
Qwen3-14B
- 10.0
Qwen3-8B
- 8.40 GLM-4.6V
- 8.20 o3-mini
- 7.80
Qwen3-4B
- 5.80
QwQ-32B
- 5.20
Qwen3-32B
- 4.00
DeepSeek-R1-Distill-Llama-70B
- 3.80
Llama 4 Scout
- 3.60
DeepSeek-R1-Distill-Qwen-32B
- 1.60 GPT-4.1
- 1.20
Gemma 3 27B
- 1.20
Mistral Small 3.2
- 0.40
Qwen2.5-72B
- 0.40
Qwen3-0.6B
- 1 GPT-6.1 Sol single-shot, medium 83.6%
- 2
DeepSeek V4.1 Flash single-shot, medium 41.8%
- 3 gpt-oss-120b single-shot 21.6%
- 4
Gemma 4 31B IT single-shot 16.6%
- 5 o4-mini single-shot 15.8%
- 6
Qwen3-14B single-shot 12.6%
- 7
Qwen3-8B single-shot 10%
- 8 GLM-4.6V single-shot 8.40%
- 9 o3-mini single-shot 8.20%
- 10
Qwen3-4B single-shot 7.80%
- 11
QwQ-32B single-shot 5.80%
- 12
Qwen3-32B single-shot 5.20%
- 13
DeepSeek-R1-Distill-Llama-70B single-shot 4.00%
- 14
Llama 4 Scout single-shot 3.80%
- 15
DeepSeek-R1-Distill-Qwen-32B single-shot 3.60%
- 16 GPT-4.1 single-shot 1.60%
- 17
Gemma 3 27B single-shot 1.20%
- 18
Mistral Small 3.2 single-shot 1.20%
- 19
Qwen2.5-72B single-shot 0.40%
- 20
Qwen3-0.6B single-shot 0.40%
Source: Official Spatial Pathfinding Reasoning Challenge leaderboard
| # | Entry | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | GPT-6.1 Sol | 83.6% | single-shot, medium, sparc (version 1-A) | Sep 2026 | Oct 2026 | Official leaderboard |
| 2 | | 41.8% | single-shot, medium, sparc (version 1-A) | Sep 2026 | Oct 2026 | Official leaderboard |
| 3 | Qwen 3.5 397B (AWQ) | 38% | single-shot, Existing GippLab run | Apr 2026 | Official leaderboard | |
| 4 | gpt-oss-120b | 21.6% | single-shot, Spatial-Gym | Aug 2025 | Apr 2026 | Official leaderboard |
| 5 | | 16.6% | single-shot, Existing GippLab run | Apr 2026 | Apr 2026 | Official leaderboard |
| 6 | OLMo 3.1 32B | 16.2% | single-shot, Spatial-Gym | Apr 2026 | Official leaderboard | |
| 7 | o4-mini | 15.8% | single-shot, SPaRC | Apr 2025 | May 2025 | Official leaderboard |
| 8 | | 12.6% | single-shot, Existing GippLab run | Apr 2025 | Aug 2025 | Official leaderboard |
| 9 | Nemotron 49B | 11.4% | single-shot, Spatial-Gym | Apr 2026 | Official leaderboard | |
| 10 | | 10% | single-shot, Existing GippLab run | Apr 2025 | Nov 2025 | Official leaderboard |
| 11 | GLM-4.6V | 8.40% | single-shot, Existing GippLab run | Dec 2025 | Apr 2026 | Official leaderboard |
| 12 | o3-mini | 8.20% | single-shot, SPaRC | Jan 2025 | May 2025 | Official leaderboard |
| 13 | | 7.80% | single-shot, Existing GippLab run | Apr 2025 | Nov 2025 | Official leaderboard |
| 14 | | 5.80% | single-shot, SPaRC | Mar 2025 | May 2025 | Official leaderboard |
| 15 | | 5.20% | single-shot, Spatial-Gym | Apr 2025 | Apr 2026 | Official leaderboard |
| 16 | | 4.00% | single-shot, SPaRC | Jan 2025 | May 2025 | Official leaderboard |
| 17 | | 3.80% | single-shot, Existing GippLab run | Apr 2025 | Apr 2026 | Official leaderboard |
| 18 | | 3.60% | single-shot, Spatial-Gym | Jan 2025 | Apr 2026 | Official leaderboard |
| 19 | GPT-4.1 | 1.60% | single-shot, SPaRC | Apr 2025 | May 2025 | Official leaderboard |
| 20 | Magistral Small | 1.60% | single-shot, Spatial-Gym | Apr 2026 | Official leaderboard | |
| 21 | | 1.20% | single-shot, SPaRC | Mar 2025 | May 2025 | Official leaderboard |
| 22 | | 1.20% | single-shot, Existing GippLab run | Jun 2025 | Apr 2026 | Official leaderboard |
| 23 | | 0.40% | single-shot, SPaRC | Sep 2024 | May 2025 | Official leaderboard |
| 24 | | 0.40% | single-shot, Spatial-Gym | Apr 2025 | Apr 2026 | Official leaderboard |
| # | Model | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | GPT-6.1 Sol | 83.6% | single-shot, medium, sparc (version 1-A) | Sep 2026 | Oct 2026 | Official leaderboard |
| 2 | | 41.8% | single-shot, medium, sparc (version 1-A) | Sep 2026 | Oct 2026 | Official leaderboard |
| 3 | gpt-oss-120b | 21.6% | single-shot, Spatial-Gym | Aug 2025 | Apr 2026 | Official leaderboard |
| 4 | | 16.6% | single-shot, Existing GippLab run | Apr 2026 | Apr 2026 | Official leaderboard |
| 5 | o4-mini | 15.8% | single-shot, SPaRC | Apr 2025 | May 2025 | Official leaderboard |
| 6 | | 12.6% | single-shot, Existing GippLab run | Apr 2025 | Aug 2025 | Official leaderboard |
| 7 | | 10% | single-shot, Existing GippLab run | Apr 2025 | Nov 2025 | Official leaderboard |
| 8 | GLM-4.6V | 8.40% | single-shot, Existing GippLab run | Dec 2025 | Apr 2026 | Official leaderboard |
| 9 | o3-mini | 8.20% | single-shot, SPaRC | Jan 2025 | May 2025 | Official leaderboard |
| 10 | | 7.80% | single-shot, Existing GippLab run | Apr 2025 | Nov 2025 | Official leaderboard |
| 11 | | 5.80% | single-shot, SPaRC | Mar 2025 | May 2025 | Official leaderboard |
| 12 | | 5.20% | single-shot, Spatial-Gym | Apr 2025 | Apr 2026 | Official leaderboard |
| 13 | | 4.00% | single-shot, SPaRC | Jan 2025 | May 2025 | Official leaderboard |
| 14 | | 3.80% | single-shot, Existing GippLab run | Apr 2025 | Apr 2026 | Official leaderboard |
| 15 | | 3.60% | single-shot, Spatial-Gym | Jan 2025 | Apr 2026 | Official leaderboard |
| 16 | GPT-4.1 | 1.60% | single-shot, SPaRC | Apr 2025 | May 2025 | Official leaderboard |
| 17 | | 1.20% | single-shot, SPaRC | Mar 2025 | May 2025 | Official leaderboard |
| 18 | | 1.20% | single-shot, Existing GippLab run | Jun 2025 | Apr 2026 | Official leaderboard |
| 19 | | 0.40% | single-shot, SPaRC | Sep 2024 | May 2025 | Official leaderboard |
| 20 | | 0.40% | single-shot, Spatial-Gym | Apr 2025 | Apr 2026 | Official leaderboard |
Rankings by part
We read 5 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Level 1, Level 2 and Level 3. GPT-6.1 Sol leads every one of them. Arrows show how far a model moved from its overall rank among the models we track.
Level 1
20 models we track have a score here.
- 1 GPT-6.1 Sol 100%
- 2
DeepSeek V4.1 Flash 89.5%
- 3 gpt-oss-120b 64%
- 4
Gemma 4 31B IT 51.2%
- 5 o4-mini 47.7%
- 6
Qwen3-14B 44.2%
- 7
Qwen3-8B 37.2%
- 8 GLM-4.6V 32.6%
- 9 o3-mini 29.1%
- 10
Qwen3-4B 26.7%
Level 2
20 models we track have a score here.
- 1 GPT-6.1 Sol 94.1%
- 2
DeepSeek V4.1 Flash 55.9%
- 3 gpt-oss-120b 26.3%
- 4
Gemma 4 31B IT 21.2%
- 5 o4-mini 19.5%
- 6
Qwen3-14B 13.6%
- 7
Qwen3-8B 11.9%
- 8 o3-mini ↑1 10.2%
- 9
Qwen3-4B ↑1 9.30%
- 10 GLM-4.6V ↓2 6.80%
Level 3
20 models we track have a score here.
- 1 GPT-6.1 Sol 87.6%
- 2
DeepSeek V4.1 Flash 40.5%
- 3 o4-mini ↑2 10.7%
- 4 gpt-oss-120b ↓1 9.90%
- 5
Gemma 4 31B IT ↓1 6.60%
- 6
Qwen3-14B 5.80%
- 7 GLM-4.6V ↑1 4.10%
- 8
Llama 4 Scout ↑6 3.30%
- 9
Qwen3-4B ↑1 3.30%
- 10 o3-mini ↓1 2.50%
Level 4
20 models we track have a score here.
- 1 GPT-6.1 Sol 82.6%
- 2
DeepSeek V4.1 Flash 15.1%
- 3 gpt-oss-120b 8.10%
- 4
Gemma 4 31B IT 4.70%
- 5
DeepSeek-R1-Distill-Qwen-32B ↑10 1.20%
- 6 o3-mini ↑3 1.20%
- 7
QwQ-32B ↑4 1.20%
- 8
Llama 4 Scout ↑6 1.20%
- 9 o4-mini ↓4 1.20%
- 10
Qwen3-14B ↓4 1.20%
Level 5
20 models we track have a score here.
- 1 GPT-6.1 Sol 49.4%
- 2
DeepSeek V4.1 Flash 4.50%
- 3 gpt-oss-120b 3.40%
- 4
Gemma 4 31B IT 2.20%
- 5
Gemma 3 27B ↑12 1.10%
- 6
Llama 4 Scout ↑8 1.10%
- 7 o4-mini ↓2 1.10%
- 8
Qwen3-14B ↓2 1.10%
- 9
Qwen2.5-72B ↑10 0%
- 10
DeepSeek-R1-Distill-Llama-70B ↑3 0%
4 models have held number one on the official leaderboard since its release in May 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Spatial Pathfinding Reasoning Challenge leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Spatial Pathfinding Reasoning Challenge 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 Games and long tasks
- ARC-AGI-3 In use Interactive game environments where agents must explore, infer goals and rules, and act efficiently without instructions. GPT-6.2 Astra 100%
- Orak Fading LLM agents play 12 popular video games across most genres through an MCP interface. Each game scores in-game progress or outcomes, and the board ranks models by their average rank across the games.
Gemini 2.5 Pro 3.90
- MineDojo Fading Agents complete open-ended tasks in Minecraft, such as mining blocks and fighting mobs, from a large simulation suite. Programmatic checks score many tasks.
Seed2.0 Pro 49%
- TextQuests Only in system cards Agents play 25 classic interactive fiction games. The score is harmful actions per 100 steps, where lower is better. The board also reports game progress.
Llama 4 Scout 7.00
- GameWorld Only in system cards Multimodal game agents play 34 browser games across 170 tasks. Outcomes are checked from the game state. GPT-6.1 Terra 76.8%
- Swarm Bench Only in system cards Models orchestrate agent swarms on complex tasks that benefit from coordinated decomposition and parallel execution. Kimi K3 76.3%
Data audit
All 10 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the official leaderboard (24 entries).
- Pass Models scored 20 models we track, plus 4 other entries on its leaderboard.
- Pass Newest score The newest is from September 2026, for GPT-6.1 Sol.
- Note Frontier models Only 1 of the top 10 has a score. The strongest tested is GPT-6.1 Sol (#6).
- Pass Headroom The best is 83.6%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on sparc.gipplab.org: 23 rows for models we track and 4 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.