Axis Spatial Spatial Intelligence Index

Search modelsMenu

Place and coordinate tasks benchmark leaderboard

Finding places by name and coordinates, and the distance and direction between them, from Place and coordinate tasks. A standalone evaluation: its scores are shown here and on model pages, but it is not counted in any index or ranking until more models have run it.

Score

All

Place and coordinate tasks score

Share of tasks passed, 0 to 100 · Higher is better

Place and coordinate tasks scoreScores 0 to 100: DeepSeek V4.1 Flash (max), Terminus-2: 100; MiMo V2.6 Flash (Reasoning), Terminus-2: 100; MiniMax M3 (Reasoning), Terminus-2: 100; Muse Glimmer 30B (high), Terminus-2: 100; Nemotron 3 Ultra (high), Terminus-2: 100; Step 5 Preview (Reasoning), Terminus-2: 100; Claude Opus 5.5 (high), Claude Code: 100; Claude Sonnet 5.5 (high), Claude Code: 100; Claude Haiku 5.5 (high), Claude Code: 100; Mistral Large 4 (high), Terminus-2: 98; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 83; Laguna XS 2.1 (Reasoning), Terminus-2: 83; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 600255075100DeepSeek V4.1 Flash (max)Terminus-2100MiMo V2.6 Flash (Reasoning)Terminus-2100MiniMax M3 (Reasoning)Terminus-2100Muse Glimmer 30B (high)Terminus-2100Nemotron 3 Ultra (high)Terminus-2100Step 5 Preview (Reasoning)Terminus-2100Claude Opus 5.5 (high)Claude Code100Claude Sonnet 5.5 (high)Claude Code100Claude Haiku 5.5 (high)Claude Code100Mistral Large 4 (high)Terminus-298Nemotron 3.5 Lightning 30B (Reasoning)Terminus-283Laguna XS 2.1 (Reasoning)Terminus-283Nemotron 3 Nano Omni 30B (Reasoning)Terminus-260
How to read this chart

What this metric means. The share of tasks the entry passed, 0 to 100.

Data table: Place and coordinate tasks score

Open weights

Place and coordinate tasks score

Share of tasks passed, 0 to 100 · Higher is better

Place and coordinate tasks scoreScores 0 to 100: DeepSeek V4.1 Flash (max), Terminus-2: 100; MiMo V2.6 Flash (Reasoning), Terminus-2: 100; MiniMax M3 (Reasoning), Terminus-2: 100; Muse Glimmer 30B (high), Terminus-2: 100; Nemotron 3 Ultra (high), Terminus-2: 100; Step 5 Preview (Reasoning), Terminus-2: 100; Mistral Large 4 (high), Terminus-2: 98; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 83; Laguna XS 2.1 (Reasoning), Terminus-2: 83; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 600255075100DeepSeek V4.1 Flash (max)Terminus-2100MiMo V2.6 Flash (Reasoning)Terminus-2100MiniMax M3 (Reasoning)Terminus-2100Muse Glimmer 30B (high)Terminus-2100Nemotron 3 Ultra (high)Terminus-2100Step 5 Preview (Reasoning)Terminus-2100Mistral Large 4 (high)Terminus-298Nemotron 3.5 Lightning 30B (Reasoning)Terminus-283Laguna XS 2.1 (Reasoning)Terminus-283Nemotron 3 Nano Omni 30B (Reasoning)Terminus-260
How to read this chart

What this metric means. The share of tasks the entry passed, 0 to 100.

Data table: Place and coordinate tasks score

All

Place and coordinate tasks: score vs. cost per task

Place and coordinate tasks score against average cost per task (USD) · Higher and further left is better

  • Most attractive quadrant
  • Pareto line
Place and coordinate tasks: score vs. cost per taskPlace and coordinate tasks score (0 to 100) against cost per task in US dollars, log scale: DeepSeek V4.1 Flash (max), Terminus-2: 100, $0.0044; Laguna XS 2.1 (Reasoning), Terminus-2: 82.93, $0.010; MiMo V2.6 Flash (Reasoning), Terminus-2: 100, $0.0017; MiniMax M3 (Reasoning), Terminus-2: 100, $0.0062; Mistral Large 4 (high), Terminus-2: 97.87, $0.0093; Muse Glimmer 30B (high), Terminus-2: 100, $0.018; Nemotron 3 Ultra (high), Terminus-2: 100, $0.013; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 83.33, $0.0052; Step 5 Preview (Reasoning), Terminus-2: 100, $0.024; Claude Opus 5.5 (high), Claude Code: 100, $0.063; Claude Sonnet 5.5 (high), Claude Code: 100, $0.030; Claude Haiku 5.5 (high), Claude Code: 100, $0.0028Most attractive quadrant80828486889092949698100$0.001$0.002$0.005$0.01$0.02$0.05$0.10Cost per task (USD, log scale)Place and coordinate tasks scoreDeepSeek V4.1 Flash (max), Terminus-2MiMo V2.6 Flash (Reasoning), Terminus-2MiniMax M3 (Reasoning), Terminus-2Muse Glimmer 30B (high), Terminus-2Nemotron 3 Ultra (high), Terminus-2Step 5 Preview (Reasoning), Terminus-2Claude Opus 5.5 (high), Claude CodeClaude Sonnet 5.5 (high), Claude CodeClaude Haiku 5.5 (high), Claude CodeMistral Large 4 (high), Terminus-2Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2Laguna XS 2.1 (Reasoning), Terminus-2
How to read this chart

What cost is measuring. What one attempt at a task would cost if paid token by token at the provider's published API prices, averaged over attempts: ordinary input tokens, input tokens served from the provider's cache at the lower cached price, tokens written to the cache where the provider charges for that, and output tokens (thinking included). Cost assumes each harness uses the model's pay-per-token API, priced at public list rates; sandboxes and our own engineering time are not included. Failed attempts count too.

How to read this chart. Each point is one entry. Further left means a lower average cost per task (log scale); higher means a higher Place and coordinate tasks score. Both axes fit the points shown, so the score axis does not start at zero. The shaded most attractive quadrant is the top-left quarter of the plot, split at the middle of each axis: a reading aid, not a fixed bar. The dotted Pareto line joins the points that no cheaper one beats on score. Only exact costs are plotted; the table lists every entry.

Data table: Place and coordinate tasks: score vs. cost per task

Open weights

Place and coordinate tasks: score vs. cost per task

Place and coordinate tasks score against average cost per task (USD) · Higher and further left is better

  • Most attractive quadrant
  • Pareto line
Place and coordinate tasks: score vs. cost per taskPlace and coordinate tasks score (0 to 100) against cost per task in US dollars, log scale: DeepSeek V4.1 Flash (max), Terminus-2: 100, $0.0044; Laguna XS 2.1 (Reasoning), Terminus-2: 82.93, $0.010; MiMo V2.6 Flash (Reasoning), Terminus-2: 100, $0.0017; MiniMax M3 (Reasoning), Terminus-2: 100, $0.0062; Mistral Large 4 (high), Terminus-2: 97.87, $0.0093; Muse Glimmer 30B (high), Terminus-2: 100, $0.018; Nemotron 3 Ultra (high), Terminus-2: 100, $0.013; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 83.33, $0.0052; Step 5 Preview (Reasoning), Terminus-2: 100, $0.024Most attractive quadrant80828486889092949698100$0.001$0.002$0.005$0.01$0.02$0.05Cost per task (USD, log scale)Place and coordinate tasks scoreDeepSeek V4.1 Flash (max), Terminus-2MiMo V2.6 Flash (Reasoning), Terminus-2MiniMax M3 (Reasoning), Terminus-2Muse Glimmer 30B (high), Terminus-2Nemotron 3 Ultra (high), Terminus-2Step 5 Preview (Reasoning), Terminus-2Mistral Large 4 (high), Terminus-2Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2Laguna XS 2.1 (Reasoning), Terminus-2
How to read this chart

What cost is measuring. What one attempt at a task would cost if paid token by token at the provider's published API prices, averaged over attempts: ordinary input tokens, input tokens served from the provider's cache at the lower cached price, tokens written to the cache where the provider charges for that, and output tokens (thinking included). Cost assumes each harness uses the model's pay-per-token API, priced at public list rates; sandboxes and our own engineering time are not included. Failed attempts count too.

How to read this chart. Each point is one entry. Further left means a lower average cost per task (log scale); higher means a higher Place and coordinate tasks score. Both axes fit the points shown, so the score axis does not start at zero. The shaded most attractive quadrant is the top-left quarter of the plot, split at the middle of each axis: a reading aid, not a fixed bar. The dotted Pareto line joins the points that no cheaper one beats on score. Only exact costs are plotted; the table lists every entry.

Data table: Place and coordinate tasks: score vs. cost per task

Token usage

All

Place and coordinate tasks: tokens per task

Average tokens per task by type · Lower is better

  • Uncached input
  • Cached input
  • Output (reasoning included)
Place and coordinate tasks: tokens per taskDeepSeek V4.1 Flash (max), Terminus-2: 18k; Mistral Large 4 (high), Terminus-2: 19k; Nemotron 3 Ultra (high), Terminus-2: 20k; MiniMax M3 (Reasoning), Terminus-2: 26k; MiMo V2.6 Flash (Reasoning), Terminus-2: 29k; Step 5 Preview (Reasoning), Terminus-2: 49k; Claude Sonnet 5.5 (high), Claude Code: 49k; Claude Opus 5.5 (high), Claude Code: 49k; Muse Glimmer 30B (high), Terminus-2: 53k; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 91k; Claude Haiku 5.5 (high), Claude Code: 99k; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 114k; Laguna XS 2.1 (Reasoning), Terminus-2: 178k050k100k150k200kDeepSeek V4.1 Flash (max)Terminus-218kMistral Large 4 (high)Terminus-219kNemotron 3 Ultra (high)Terminus-220kMiniMax M3 (Reasoning)Terminus-226kMiMo V2.6 Flash (Reasoning)Terminus-229kStep 5 Preview (Reasoning)Terminus-249kClaude Sonnet 5.5 (high)Claude Code49kClaude Opus 5.5 (high)Claude Code49kMuse Glimmer 30B (high)Terminus-253kNemotron 3.5 Lightning 30B (Reasoning)Terminus-291kClaude Haiku 5.5 (high)Claude Code99kNemotron 3 Nano Omni 30B (Reasoning)Terminus-2114kLaguna XS 2.1 (Reasoning)Terminus-2178k
How to read this chart

What the parts mean. Uncached input: input tokens read at the full price. Cached input: input tokens served from the provider's cache at the lower cached price. Output: the tokens the model wrote, its reasoning included, as reported by the provider.

Compare with care. Models use different tokenisers, so the same text is a different number of tokens for different models. Use token counts to see how much work each model did, and cost to compare spending.

Data table: Place and coordinate tasks: tokens per task
Place and coordinate tasks: tokens per task
ModelDeveloperUncached inputCached inputOutput (reasoning included)Total
DeepSeek V4.1 Flash (max), Terminus-2DeepSeek5k10k2k18k
Mistral Large 4 (high), Terminus-2Mistral AI5k11k2k19k
Nemotron 3 Ultra (high), Terminus-2NVIDIA14k4k2k20k
MiniMax M3 (Reasoning), Terminus-2MiniMax7k16k3k26k
MiMo V2.6 Flash (Reasoning), Terminus-2Xiaomi7k20k2k29k
Step 5 Preview (Reasoning), Terminus-2StepFun4k40k5k49k
Claude Sonnet 5.5 (high), Claude CodeAnthropic5k44k52349k
Claude Opus 5.5 (high), Claude CodeAnthropic5k43k58149k
Muse Glimmer 30B (high), Terminus-2Meta34k15k4k53k
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2NVIDIA23k63k5k91k
Claude Haiku 5.5 (high), Claude CodeAnthropic6k92k1k99k
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2NVIDIA15k90k10k114k
Laguna XS 2.1 (Reasoning), Terminus-2Poolside12k159k6k178k

Open weights

Place and coordinate tasks: tokens per task

Average tokens per task by type · Lower is better

  • Uncached input
  • Cached input
  • Output (reasoning included)
Place and coordinate tasks: tokens per taskDeepSeek V4.1 Flash (max), Terminus-2: 18k; Mistral Large 4 (high), Terminus-2: 19k; Nemotron 3 Ultra (high), Terminus-2: 20k; MiniMax M3 (Reasoning), Terminus-2: 26k; MiMo V2.6 Flash (Reasoning), Terminus-2: 29k; Step 5 Preview (Reasoning), Terminus-2: 49k; Muse Glimmer 30B (high), Terminus-2: 53k; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 91k; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 114k; Laguna XS 2.1 (Reasoning), Terminus-2: 178k050k100k150k200kDeepSeek V4.1 Flash (max)Terminus-218kMistral Large 4 (high)Terminus-219kNemotron 3 Ultra (high)Terminus-220kMiniMax M3 (Reasoning)Terminus-226kMiMo V2.6 Flash (Reasoning)Terminus-229kStep 5 Preview (Reasoning)Terminus-249kMuse Glimmer 30B (high)Terminus-253kNemotron 3.5 Lightning 30B (Reasoning)Terminus-291kNemotron 3 Nano Omni 30B (Reasoning)Terminus-2114kLaguna XS 2.1 (Reasoning)Terminus-2178k
How to read this chart

What the parts mean. Uncached input: input tokens read at the full price. Cached input: input tokens served from the provider's cache at the lower cached price. Output: the tokens the model wrote, its reasoning included, as reported by the provider.

Compare with care. Models use different tokenisers, so the same text is a different number of tokens for different models. Use token counts to see how much work each model did, and cost to compare spending.

Data table: Place and coordinate tasks: tokens per task
Place and coordinate tasks: tokens per task
ModelDeveloperUncached inputCached inputOutput (reasoning included)Total
DeepSeek V4.1 Flash (max), Terminus-2DeepSeek5k10k2k18k
Mistral Large 4 (high), Terminus-2Mistral AI5k11k2k19k
Nemotron 3 Ultra (high), Terminus-2NVIDIA14k4k2k20k
MiniMax M3 (Reasoning), Terminus-2MiniMax7k16k3k26k
MiMo V2.6 Flash (Reasoning), Terminus-2Xiaomi7k20k2k29k
Step 5 Preview (Reasoning), Terminus-2StepFun4k40k5k49k
Muse Glimmer 30B (high), Terminus-2Meta34k15k4k53k
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2NVIDIA23k63k5k91k
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2NVIDIA15k90k10k114k
Laguna XS 2.1 (Reasoning), Terminus-2Poolside12k159k6k178k

Cost

All

Place and coordinate tasks: cost per task

Average cost per task (USD) by token type · Lower is better

  • Uncached input
  • Cached input
  • Output
  • Not split by token type
Place and coordinate tasks: cost per taskValues in US dollars: MiMo V2.6 Flash (Reasoning), Terminus-2: $0.0017; Claude Haiku 5.5 (high), Claude Code: $0.0028; DeepSeek V4.1 Flash (max), Terminus-2: $0.0044; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: $0.0052; MiniMax M3 (Reasoning), Terminus-2: $0.0062; Mistral Large 4 (high), Terminus-2: $0.0093; Laguna XS 2.1 (Reasoning), Terminus-2: $0.010; Nemotron 3 Ultra (high), Terminus-2: $0.013; Muse Glimmer 30B (high), Terminus-2: $0.018; Step 5 Preview (Reasoning), Terminus-2: $0.024; Claude Sonnet 5.5 (high), Claude Code: $0.030; Claude Opus 5.5 (high), Claude Code: $0.063$0$0.02$0.04$0.06$0.08MiMo V2.6 Flash (Reasoning)Terminus-2$0.0017Claude Haiku 5.5 (high)Claude Code$0.0028DeepSeek V4.1 Flash (max)Terminus-2$0.0044Nemotron 3.5 Lightning 30B (Reasoning)Terminus-2$0.0052MiniMax M3 (Reasoning)Terminus-2$0.0062Mistral Large 4 (high)Terminus-2$0.0093Laguna XS 2.1 (Reasoning)Terminus-2$0.010Nemotron 3 Ultra (high)Terminus-2$0.013Muse Glimmer 30B (high)Terminus-2$0.018Step 5 Preview (Reasoning)Terminus-2$0.024Claude Sonnet 5.5 (high)Claude Code$0.030Claude Opus 5.5 (high)Claude Code$0.063
How to read this chart

What cost is measuring. What one attempt at a task would cost if paid token by token at the provider's published API prices, averaged over attempts: ordinary input tokens, input tokens served from the provider's cache at the lower cached price, tokens written to the cache where the provider charges for that, and output tokens (thinking included). Cost assumes each harness uses the model's pay-per-token API, priced at public list rates; sandboxes and our own engineering time are not included. Failed attempts count too.

How the cost is split. Each part is the entry's average tokens of that type times its list price. The rest, mostly tokens written to the cache, is shown as not split by token type.

Data table: Place and coordinate tasks: cost per task
Place and coordinate tasks: cost per task
ModelDeveloperUncached inputCached inputOutputNot split by token typeTotal
MiMo V2.6 Flash (Reasoning), Terminus-2Xiaomi$0.00093$0.000056$0.00069$0.00$0.0017
Claude Haiku 5.5 (high), Claude CodeAnthropic$0.00063$0.00092$0.00058$0.00063$0.0028
DeepSeek V4.1 Flash (max), Terminus-2DeepSeek$0.00077$0.000031$0.0014$0.0022$0.0044
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2NVIDIA$0.0014$0.0028$0.0011$0.00000027$0.0052
MiniMax M3 (Reasoning), Terminus-2MiniMax$0.0021$0.00097$0.0031$0.00000027$0.0062
Mistral Large 4 (high), Terminus-2Mistral AI$0.0033$0.00079$0.0052$0.00$0.0093
Laguna XS 2.1 (Reasoning), Terminus-2Poolside$0.0012$0.0080$0.0012$0.00$0.010
Nemotron 3 Ultra (high), Terminus-2NVIDIA$0.0084$0.00044$0.0047$0.00$0.013
Muse Glimmer 30B (high), Terminus-2Meta$0.012$0.00061$0.0057$0.00$0.018
Step 5 Preview (Reasoning), Terminus-2StepFun$0.0043$0.0042$0.015$0.00000013$0.024
Claude Sonnet 5.5 (high), Claude CodeAnthropic$0.010$0.0044$0.0052$0.0099$0.030
Claude Opus 5.5 (high), Claude CodeAnthropic$0.021$0.0087$0.012$0.021$0.063
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2NVIDIANo dataNo dataNo dataNo dataNo data

Open weights

Place and coordinate tasks: cost per task

Average cost per task (USD) by token type · Lower is better

  • Uncached input
  • Cached input
  • Output
  • Not split by token type
Place and coordinate tasks: cost per taskValues in US dollars: MiMo V2.6 Flash (Reasoning), Terminus-2: $0.0017; DeepSeek V4.1 Flash (max), Terminus-2: $0.0044; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: $0.0052; MiniMax M3 (Reasoning), Terminus-2: $0.0062; Mistral Large 4 (high), Terminus-2: $0.0093; Laguna XS 2.1 (Reasoning), Terminus-2: $0.010; Nemotron 3 Ultra (high), Terminus-2: $0.013; Muse Glimmer 30B (high), Terminus-2: $0.018; Step 5 Preview (Reasoning), Terminus-2: $0.024$0$0.01$0.02$0.03$0.04MiMo V2.6 Flash (Reasoning)Terminus-2$0.0017DeepSeek V4.1 Flash (max)Terminus-2$0.0044Nemotron 3.5 Lightning 30B (Reasoning)Terminus-2$0.0052MiniMax M3 (Reasoning)Terminus-2$0.0062Mistral Large 4 (high)Terminus-2$0.0093Laguna XS 2.1 (Reasoning)Terminus-2$0.010Nemotron 3 Ultra (high)Terminus-2$0.013Muse Glimmer 30B (high)Terminus-2$0.018Step 5 Preview (Reasoning)Terminus-2$0.024
How to read this chart

What cost is measuring. What one attempt at a task would cost if paid token by token at the provider's published API prices, averaged over attempts: ordinary input tokens, input tokens served from the provider's cache at the lower cached price, tokens written to the cache where the provider charges for that, and output tokens (thinking included). Cost assumes each harness uses the model's pay-per-token API, priced at public list rates; sandboxes and our own engineering time are not included. Failed attempts count too.

How the cost is split. Each part is the entry's average tokens of that type times its list price. The rest, mostly tokens written to the cache, is shown as not split by token type.

Data table: Place and coordinate tasks: cost per task
Place and coordinate tasks: cost per task
ModelDeveloperUncached inputCached inputOutputNot split by token typeTotal
MiMo V2.6 Flash (Reasoning), Terminus-2Xiaomi$0.00093$0.000056$0.00069$0.00$0.0017
DeepSeek V4.1 Flash (max), Terminus-2DeepSeek$0.00077$0.000031$0.0014$0.0022$0.0044
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2NVIDIA$0.0014$0.0028$0.0011$0.00000027$0.0052
MiniMax M3 (Reasoning), Terminus-2MiniMax$0.0021$0.00097$0.0031$0.00000027$0.0062
Mistral Large 4 (high), Terminus-2Mistral AI$0.0033$0.00079$0.0052$0.00$0.0093
Laguna XS 2.1 (Reasoning), Terminus-2Poolside$0.0012$0.0080$0.0012$0.00$0.010
Nemotron 3 Ultra (high), Terminus-2NVIDIA$0.0084$0.00044$0.0047$0.00$0.013
Muse Glimmer 30B (high), Terminus-2Meta$0.012$0.00061$0.0057$0.00$0.018
Step 5 Preview (Reasoning), Terminus-2StepFun$0.0043$0.0042$0.015$0.00000013$0.024
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2NVIDIANo dataNo dataNo dataNo dataNo data

Speed

All

Place and coordinate tasks: time per task

Time per task (at 100 output tokens/s), minutes · Lower is better

Place and coordinate tasks: time per taskPlace and coordinate tasks: time per task: Claude Sonnet 5.5 (high), Claude Code: 0.2 min; Claude Opus 5.5 (high), Claude Code: 0.2 min; Claude Haiku 5.5 (high), Claude Code: 0.3 min; Nemotron 3 Ultra (high), Terminus-2: 0.4 min; DeepSeek V4.1 Flash (max), Terminus-2: 0.5 min; Mistral Large 4 (high), Terminus-2: 0.7 min; MiniMax M3 (Reasoning), Terminus-2: 0.7 min; Muse Glimmer 30B (high), Terminus-2: 0.8 min; MiMo V2.6 Flash (Reasoning), Terminus-2: 0.8 min; Step 5 Preview (Reasoning), Terminus-2: 1.2 min; Laguna XS 2.1 (Reasoning), Terminus-2: 1.3 min; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 1.4 min; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 2.2 min0 min1 min2 min3 min4 minClaude Sonnet 5.5 (high)Claude Code0.2 minClaude Opus 5.5 (high)Claude Code0.2 minClaude Haiku 5.5 (high)Claude Code0.3 minNemotron 3 Ultra (high)Terminus-20.4 minDeepSeek V4.1 Flash (max)Terminus-20.5 minMistral Large 4 (high)Terminus-20.7 minMiniMax M3 (Reasoning)Terminus-20.7 minMuse Glimmer 30B (high)Terminus-20.8 minMiMo V2.6 Flash (Reasoning)Terminus-20.8 minStep 5 Preview (Reasoning)Terminus-21.2 minLaguna XS 2.1 (Reasoning)Terminus-21.3 minNemotron 3.5 Lightning 30B (Reasoning)Terminus-21.4 minNemotron 3 Nano Omni 30B (Reasoning)Terminus-22.2 min
How to read this chart

Time per task. Time as if every model wrote 100 output tokens per second: output tokens divided by 100, plus the time the agent's commands took. Serving speed differs between providers and is not ranked.

Data table: Place and coordinate tasks: time per task

Open weights

Place and coordinate tasks: time per task

Time per task (at 100 output tokens/s), minutes · Lower is better

Place and coordinate tasks: time per taskPlace and coordinate tasks: time per task: Nemotron 3 Ultra (high), Terminus-2: 0.4 min; DeepSeek V4.1 Flash (max), Terminus-2: 0.5 min; Mistral Large 4 (high), Terminus-2: 0.7 min; MiniMax M3 (Reasoning), Terminus-2: 0.7 min; Muse Glimmer 30B (high), Terminus-2: 0.8 min; MiMo V2.6 Flash (Reasoning), Terminus-2: 0.8 min; Step 5 Preview (Reasoning), Terminus-2: 1.2 min; Laguna XS 2.1 (Reasoning), Terminus-2: 1.3 min; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 1.4 min; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 2.2 min0 min1 min2 min3 min4 minNemotron 3 Ultra (high)Terminus-20.4 minDeepSeek V4.1 Flash (max)Terminus-20.5 minMistral Large 4 (high)Terminus-20.7 minMiniMax M3 (Reasoning)Terminus-20.7 minMuse Glimmer 30B (high)Terminus-20.8 minMiMo V2.6 Flash (Reasoning)Terminus-20.8 minStep 5 Preview (Reasoning)Terminus-21.2 minLaguna XS 2.1 (Reasoning)Terminus-21.3 minNemotron 3.5 Lightning 30B (Reasoning)Terminus-21.4 minNemotron 3 Nano Omni 30B (Reasoning)Terminus-22.2 min
How to read this chart

Time per task. Time as if every model wrote 100 output tokens per second: output tokens divided by 100, plus the time the agent's commands took. Serving speed differs between providers and is not ranked.

Data table: Place and coordinate tasks: time per task

All

Place and coordinate tasks: score vs. time per task

Place and coordinate tasks score against time per task (at 100 output tokens/s), minutes · Higher and further left is better

  • Most attractive quadrant
  • Pareto line
Place and coordinate tasks: score vs. time per taskPlace and coordinate tasks score (0 to 100) against time per task (minutes), log scale: DeepSeek V4.1 Flash (max), Terminus-2: 100, 0.5 min; Laguna XS 2.1 (Reasoning), Terminus-2: 82.93, 1.3 min; MiMo V2.6 Flash (Reasoning), Terminus-2: 100, 0.8 min; MiniMax M3 (Reasoning), Terminus-2: 100, 0.7 min; Mistral Large 4 (high), Terminus-2: 97.87, 0.7 min; Muse Glimmer 30B (high), Terminus-2: 100, 0.8 min; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 60.47, 2.2 min; Nemotron 3 Ultra (high), Terminus-2: 100, 0.4 min; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 83.33, 1.4 min; Step 5 Preview (Reasoning), Terminus-2: 100, 1.2 min; Claude Opus 5.5 (high), Claude Code: 100, 0.2 min; Claude Sonnet 5.5 (high), Claude Code: 100, 0.2 min; Claude Haiku 5.5 (high), Claude Code: 100, 0.3 minMost attractive quadrant5560657075808590951000.1 min0.2 min0.5 min1 min2 min5 minTime per task (minutes, log scale)Place and coordinate tasks scoreDeepSeek V4.1 Flash (max), Terminus-2MiniMax M3 (Reasoning), Terminus-2Nemotron 3 Ultra (high), Terminus-2Claude Sonnet 5.5 (high), Claude CodeMiMo V2.6 Flash (Reasoning), Terminus-2Muse Glimmer 30B (high), Terminus-2Step 5 Preview (Reasoning), Terminus-2Claude Opus 5.5 (high), Claude CodeClaude Haiku 5.5 (high), Claude CodeMistral Large 4 (high), Terminus-2Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2Laguna XS 2.1 (Reasoning), Terminus-2Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2
How to read this chart

Time per task. Time as if every model wrote 100 output tokens per second: output tokens divided by 100, plus the time the agent's commands took. Serving speed differs between providers and is not ranked.

How to read this chart. Each point is one entry. Further left means a lower average time per task (log scale); higher means a higher Place and coordinate tasks score. Both axes fit the points shown, so the score axis does not start at zero. The shaded most attractive quadrant is the top-left quarter of the plot: a reading aid, not a fixed bar. The dotted Pareto line joins the points that no faster one beats on score.

Data table: Place and coordinate tasks: score vs. time per task

Open weights

Place and coordinate tasks: score vs. time per task

Place and coordinate tasks score against time per task (at 100 output tokens/s), minutes · Higher and further left is better

  • Most attractive quadrant
  • Pareto line
Place and coordinate tasks: score vs. time per taskPlace and coordinate tasks score (0 to 100) against time per task (minutes), log scale: DeepSeek V4.1 Flash (max), Terminus-2: 100, 0.5 min; Laguna XS 2.1 (Reasoning), Terminus-2: 82.93, 1.3 min; MiMo V2.6 Flash (Reasoning), Terminus-2: 100, 0.8 min; MiniMax M3 (Reasoning), Terminus-2: 100, 0.7 min; Mistral Large 4 (high), Terminus-2: 97.87, 0.7 min; Muse Glimmer 30B (high), Terminus-2: 100, 0.8 min; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 60.47, 2.2 min; Nemotron 3 Ultra (high), Terminus-2: 100, 0.4 min; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 83.33, 1.4 min; Step 5 Preview (Reasoning), Terminus-2: 100, 1.2 minMost attractive quadrant5560657075808590951000.2 min0.5 min1 min2 min5 minTime per task (minutes, log scale)Place and coordinate tasks scoreDeepSeek V4.1 Flash (max), Terminus-2MiMo V2.6 Flash (Reasoning), Terminus-2MiniMax M3 (Reasoning), Terminus-2Nemotron 3 Ultra (high), Terminus-2Muse Glimmer 30B (high), Terminus-2Step 5 Preview (Reasoning), Terminus-2Mistral Large 4 (high), Terminus-2Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2Laguna XS 2.1 (Reasoning), Terminus-2Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2
How to read this chart

Time per task. Time as if every model wrote 100 output tokens per second: output tokens divided by 100, plus the time the agent's commands took. Serving speed differs between providers and is not ranked.

How to read this chart. Each point is one entry. Further left means a lower average time per task (log scale); higher means a higher Place and coordinate tasks score. Both axes fit the points shown, so the score axis does not start at zero. The shaded most attractive quadrant is the top-left quarter of the plot: a reading aid, not a fixed bar. The dotted Pareto line joins the points that no faster one beats on score.

Data table: Place and coordinate tasks: score vs. time per task

Score vs. release date

All

Place and coordinate tasks: score vs. release date

Place and coordinate tasks score against the model's public release date

Place and coordinate tasks: score vs. release datePlace and coordinate tasks score (0 to 100) against release date: DeepSeek V4.1 Flash (max), Terminus-2: 100, 10 September 2026; Laguna XS 2.1 (Reasoning), Terminus-2: 82.93, 2 July 2026; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 60.47, 28 April 2026; Nemotron 3 Ultra (high), Terminus-2: 100, 4 June 2026; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 83.33, 11 August 2026; Claude Opus 5.5 (high), Claude Code: 100, 22 September 2026; Claude Sonnet 5.5 (high), Claude Code: 100, 28 September 2026; Claude Haiku 5.5 (high), Claude Code: 100, 7 October 2026556065707580859095100Mar 2026May 2026Jul 2026Sep 2026Nov 2026Release datePlace and coordinate tasks scoreDeepSeek V4.1 Flash (max), Terminus-2Nemotron 3 Ultra (high), Terminus-2Claude Opus 5.5 (high), Claude CodeClaude Sonnet 5.5 (high), Claude CodeClaude Haiku 5.5 (high), Claude CodeNemotron 3.5 Lightning 30B (Reasoning), Terminus-2Laguna XS 2.1 (Reasoning), Terminus-2Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2
How to read this chart

How to read this chart. Each point is one entry: across, the date its model was first released; up, its Place and coordinate tasks score. Models with no published release date are not plotted.

Data table: Place and coordinate tasks: score vs. release date

Open weights

Place and coordinate tasks: score vs. release date

Place and coordinate tasks score against the model's public release date

Place and coordinate tasks: score vs. release datePlace and coordinate tasks score (0 to 100) against release date: DeepSeek V4.1 Flash (max), Terminus-2: 100, 10 September 2026; Laguna XS 2.1 (Reasoning), Terminus-2: 82.93, 2 July 2026; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 60.47, 28 April 2026; Nemotron 3 Ultra (high), Terminus-2: 100, 4 June 2026; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 83.33, 11 August 2026556065707580859095100Apr 2026May 2026Jun 2026Jul 2026Aug 2026Sep 2026Oct 2026Release datePlace and coordinate tasks scoreDeepSeek V4.1 Flash (max), Terminus-2Nemotron 3 Ultra (high), Terminus-2Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2Laguna XS 2.1 (Reasoning), Terminus-2Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2
How to read this chart

How to read this chart. Each point is one entry: across, the date its model was first released; up, its Place and coordinate tasks score. Models with no published release date are not plotted.

Data table: Place and coordinate tasks: score vs. release date

Background

Tasks we built, in four families: find a place's coordinates from its name, name the place at a coordinate, measure the distance between two places, and give the direction from one to another. Every continent is covered in each family, with planted hard cases: places that share a name, old names, routes across the 180th meridian and long routes where a straight line on the map gives the wrong bearing.

Licence: GeoNames data CC BY 4.0; Natural Earth public domain.

Published by us, from GeoNames and Natural Earth data.

Related links

Explore evaluations