Axis Spatial Spatial Intelligence Index

Search modelsMenu

InSAR Regional Snow Mapping benchmark leaderboard

Checks over radar interferograms, weather and snow layers in the Alps, from InSAR Regional Snow Mapping. 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

InSAR Regional Snow Mapping score

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

InSAR Regional Snow Mapping scoreScores 0 to 100: Mistral Large 4 (high), Terminus-2: 100; Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2: 100; Qwen 3.8 27B (xhigh), Terminus-2: 100; Step 5 Preview (Reasoning), Terminus-2: 100; Claude Opus 5.5 (xhigh), Claude Code: 100; Claude Sonnet 5.5 (xhigh), Claude Code: 100; Claude Haiku 5.5 (xhigh), Claude Code: 100; Claude Opus 5 (high), Claude Code: 100; Claude Sonnet 5 (high), Claude Code: 100; Claude Opus 4.8 (high), Claude Code: 100; Claude Opus 4.7 (high), Claude Code: 100; Muse Glimmer 30B (high), Terminus-2: 96; Nemotron 3.5 Super 120B (Reasoning), Terminus-2: 96; Claude Haiku 4.5 (Reasoning), Claude Code: 96; Laguna XS 2.1 (Reasoning), Terminus-2: 88; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 79; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 330255075100Mistral Large 4 (high)Terminus-2100Nemotron 3.5 Super VL 120B (Reasoning)Terminus-2100Qwen 3.8 27B (xhigh)Terminus-2100Step 5 Preview (Reasoning)Terminus-2100Claude Opus 5.5 (xhigh)Claude Code100Claude Sonnet 5.5 (xhigh)Claude Code100Claude Haiku 5.5 (xhigh)Claude Code100Claude Opus 5 (high)Claude Code100Claude Sonnet 5 (high)Claude Code100Claude Opus 4.8 (high)Claude Code100Claude Opus 4.7 (high)Claude Code100Muse Glimmer 30B (high)Terminus-296Nemotron 3.5 Super 120B (Reasoning)Terminus-296Claude Haiku 4.5 (Reasoning)Claude Code96Laguna XS 2.1 (Reasoning)Terminus-288Nemotron 3.5 Lightning 30B (Reasoning)Terminus-279Nemotron 3 Nano Omni 30B (Reasoning)Terminus-233
How to read this chart

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

Data table: InSAR Regional Snow Mapping score

Open weights

InSAR Regional Snow Mapping score

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

InSAR Regional Snow Mapping scoreScores 0 to 100: Mistral Large 4 (high), Terminus-2: 100; Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2: 100; Qwen 3.8 27B (xhigh), Terminus-2: 100; Step 5 Preview (Reasoning), Terminus-2: 100; Muse Glimmer 30B (high), Terminus-2: 96; Nemotron 3.5 Super 120B (Reasoning), Terminus-2: 96; Laguna XS 2.1 (Reasoning), Terminus-2: 88; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 79; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 330255075100Mistral Large 4 (high)Terminus-2100Nemotron 3.5 Super VL 120B (Reasoning)Terminus-2100Qwen 3.8 27B (xhigh)Terminus-2100Step 5 Preview (Reasoning)Terminus-2100Muse Glimmer 30B (high)Terminus-296Nemotron 3.5 Super 120B (Reasoning)Terminus-296Laguna XS 2.1 (Reasoning)Terminus-288Nemotron 3.5 Lightning 30B (Reasoning)Terminus-279Nemotron 3 Nano Omni 30B (Reasoning)Terminus-233
How to read this chart

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

Data table: InSAR Regional Snow Mapping score

All

InSAR Regional Snow Mapping: score vs. cost per task

InSAR Regional Snow Mapping score against average cost per task (USD) · Higher and further left is better

  • Most attractive quadrant
  • Pareto line
InSAR Regional Snow Mapping: score vs. cost per taskInSAR Regional Snow Mapping score (0 to 100) against cost per task in US dollars, log scale: Laguna XS 2.1 (Reasoning), Terminus-2: 87.5, $0.037; Mistral Large 4 (high), Terminus-2: 100, $0.028; Muse Glimmer 30B (high), Terminus-2: 95.83, $0.038; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 79.17, $0.012; Nemotron 3.5 Super 120B (Reasoning), Terminus-2: 95.83, $0.0081; Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2: 100, $0.0061; Qwen 3.8 27B (xhigh), Terminus-2: 100, $0.049; Step 5 Preview (Reasoning), Terminus-2: 100, $0.093; Claude Haiku 4.5 (Reasoning), Claude Code: 95.83, $0.068; Claude Opus 5.5 (xhigh), Claude Code: 100, $0.14; Claude Sonnet 5.5 (xhigh), Claude Code: 100, $0.055; Claude Haiku 5.5 (xhigh), Claude Code: 100, $0.0079; Claude Opus 5 (high), Claude Code: 100, $0.15; Claude Sonnet 5 (high), Claude Code: 100, $0.090; Claude Opus 4.8 (high), Claude Code: 100, $0.15; Claude Opus 4.7 (high), Claude Code: 100, $0.20Most attractive quadrant7580859095100$0.005$0.01$0.02$0.05$0.10$0.20$0.50Cost per task (USD, log scale)InSAR Regional Snow Mapping scoreMistral Large 4 (high), Terminus-2Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2Qwen 3.8 27B (xhigh), Terminus-2Step 5 Preview (Reasoning), Terminus-2Claude Opus 5.5 (xhigh), Claude CodeClaude Sonnet 5.5 (xhigh), Claude CodeClaude Haiku 5.5 (xhigh), Claude CodeClaude Opus 5 (high), Claude CodeClaude Sonnet 5 (high), Claude CodeClaude Opus 4.8 (high), Claude CodeClaude Opus 4.7 (high), Claude CodeMuse Glimmer 30B (high), Terminus-2Nemotron 3.5 Super 120B (Reasoning), Terminus-2Claude Haiku 4.5 (Reasoning), Claude CodeLaguna XS 2.1 (Reasoning), Terminus-2Nemotron 3.5 Lightning 30B (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 InSAR Regional Snow Mapping 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: InSAR Regional Snow Mapping: score vs. cost per task

Open weights

InSAR Regional Snow Mapping: score vs. cost per task

InSAR Regional Snow Mapping score against average cost per task (USD) · Higher and further left is better

  • Most attractive quadrant
  • Pareto line
InSAR Regional Snow Mapping: score vs. cost per taskInSAR Regional Snow Mapping score (0 to 100) against cost per task in US dollars, log scale: Laguna XS 2.1 (Reasoning), Terminus-2: 87.5, $0.037; Mistral Large 4 (high), Terminus-2: 100, $0.028; Muse Glimmer 30B (high), Terminus-2: 95.83, $0.038; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 79.17, $0.012; Nemotron 3.5 Super 120B (Reasoning), Terminus-2: 95.83, $0.0081; Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2: 100, $0.0061; Qwen 3.8 27B (xhigh), Terminus-2: 100, $0.049; Step 5 Preview (Reasoning), Terminus-2: 100, $0.093Most attractive quadrant7580859095100$0.005$0.01$0.02$0.05$0.10$0.20Cost per task (USD, log scale)InSAR Regional Snow Mapping scoreMistral Large 4 (high), Terminus-2Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2Nemotron 3.5 Super 120B (Reasoning), Terminus-2Qwen 3.8 27B (xhigh), Terminus-2Step 5 Preview (Reasoning), Terminus-2Muse Glimmer 30B (high), Terminus-2Laguna XS 2.1 (Reasoning), Terminus-2Nemotron 3.5 Lightning 30B (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 InSAR Regional Snow Mapping 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: InSAR Regional Snow Mapping: score vs. cost per task

Token usage

All

InSAR Regional Snow Mapping: tokens per task

Average tokens per task by type · Lower is better

  • Uncached input
  • Cached input
  • Output (reasoning included)
InSAR Regional Snow Mapping: tokens per taskNemotron 3.5 Super VL 120B (Reasoning), Terminus-2: 44k; Nemotron 3.5 Super 120B (Reasoning), Terminus-2: 47k; Claude Opus 4.8 (high), Claude Code: 62k; Claude Sonnet 5.5 (xhigh), Claude Code: 68k; Mistral Large 4 (high), Terminus-2: 69k; Claude Opus 5 (high), Claude Code: 71k; Claude Opus 5.5 (xhigh), Claude Code: 86k; Qwen 3.8 27B (xhigh), Terminus-2: 93k; Claude Opus 4.7 (high), Claude Code: 128k; Step 5 Preview (Reasoning), Terminus-2: 149k; Muse Glimmer 30B (high), Terminus-2: 151k; Claude Sonnet 5 (high), Claude Code: 154k; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 208k; Claude Haiku 4.5 (Reasoning), Claude Code: 234k; Claude Haiku 5.5 (xhigh), Claude Code: 235k; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 514k; Laguna XS 2.1 (Reasoning), Terminus-2: 652k0200k400k600k800kNemotron 3.5 Super VL 120B (Reasoning)Terminus-244kNemotron 3.5 Super 120B (Reasoning)Terminus-247kClaude Opus 4.8 (high)Claude Code62kClaude Sonnet 5.5 (xhigh)Claude Code68kMistral Large 4 (high)Terminus-269kClaude Opus 5 (high)Claude Code71kClaude Opus 5.5 (xhigh)Claude Code86kQwen 3.8 27B (xhigh)Terminus-293kClaude Opus 4.7 (high)Claude Code128kStep 5 Preview (Reasoning)Terminus-2149kMuse Glimmer 30B (high)Terminus-2151kClaude Sonnet 5 (high)Claude Code154kNemotron 3.5 Lightning 30B (Reasoning)Terminus-2208kClaude Haiku 4.5 (Reasoning)Claude Code234kClaude Haiku 5.5 (xhigh)Claude Code235kNemotron 3 Nano Omni 30B (Reasoning)Terminus-2514kLaguna XS 2.1 (Reasoning)Terminus-2652k
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: InSAR Regional Snow Mapping: tokens per task
InSAR Regional Snow Mapping: tokens per task
ModelDeveloperUncached inputCached inputOutput (reasoning included)Total
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2NVIDIA24k16k4k44k
Nemotron 3.5 Super 120B (Reasoning), Terminus-2NVIDIA24k18k5k47k
Claude Opus 4.8 (high), Claude CodeAnthropic9k52k1k62k
Claude Sonnet 5.5 (xhigh), Claude CodeAnthropic8k58k2k68k
Mistral Large 4 (high), Terminus-2Mistral AI16k47k7k69k
Claude Opus 5 (high), Claude CodeAnthropic8k62k1k71k
Claude Opus 5.5 (xhigh), Claude CodeAnthropic11k73k2k86k
Qwen 3.8 27B (xhigh), Terminus-2Alibaba (Qwen)44k40k8k93k
Claude Opus 4.7 (high), Claude CodeAnthropic10k116k1k128k
Step 5 Preview (Reasoning), Terminus-2StepFun42k94k13k149k
Muse Glimmer 30B (high), Terminus-2Meta66k77k8k151k
Claude Sonnet 5 (high), Claude CodeAnthropic10k141k2k154k
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2NVIDIA39k156k12k208k
Claude Haiku 4.5 (Reasoning), Claude CodeAnthropic11k218k5k234k
Claude Haiku 5.5 (xhigh), Claude CodeAnthropic16k213k5k235k
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2NVIDIA54k441k20k514k
Laguna XS 2.1 (Reasoning), Terminus-2Poolside45k591k16k652k

Open weights

InSAR Regional Snow Mapping: tokens per task

Average tokens per task by type · Lower is better

  • Uncached input
  • Cached input
  • Output (reasoning included)
InSAR Regional Snow Mapping: tokens per taskNemotron 3.5 Super VL 120B (Reasoning), Terminus-2: 44k; Nemotron 3.5 Super 120B (Reasoning), Terminus-2: 47k; Mistral Large 4 (high), Terminus-2: 69k; Qwen 3.8 27B (xhigh), Terminus-2: 93k; Step 5 Preview (Reasoning), Terminus-2: 149k; Muse Glimmer 30B (high), Terminus-2: 151k; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 208k; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 514k; Laguna XS 2.1 (Reasoning), Terminus-2: 652k0200k400k600k800kNemotron 3.5 Super VL 120B (Reasoning)Terminus-244kNemotron 3.5 Super 120B (Reasoning)Terminus-247kMistral Large 4 (high)Terminus-269kQwen 3.8 27B (xhigh)Terminus-293kStep 5 Preview (Reasoning)Terminus-2149kMuse Glimmer 30B (high)Terminus-2151kNemotron 3.5 Lightning 30B (Reasoning)Terminus-2208kNemotron 3 Nano Omni 30B (Reasoning)Terminus-2514kLaguna XS 2.1 (Reasoning)Terminus-2652k
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: InSAR Regional Snow Mapping: tokens per task
InSAR Regional Snow Mapping: tokens per task
ModelDeveloperUncached inputCached inputOutput (reasoning included)Total
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2NVIDIA24k16k4k44k
Nemotron 3.5 Super 120B (Reasoning), Terminus-2NVIDIA24k18k5k47k
Mistral Large 4 (high), Terminus-2Mistral AI16k47k7k69k
Qwen 3.8 27B (xhigh), Terminus-2Alibaba (Qwen)44k40k8k93k
Step 5 Preview (Reasoning), Terminus-2StepFun42k94k13k149k
Muse Glimmer 30B (high), Terminus-2Meta66k77k8k151k
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2NVIDIA39k156k12k208k
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2NVIDIA54k441k20k514k
Laguna XS 2.1 (Reasoning), Terminus-2Poolside45k591k16k652k

Cost

All

InSAR Regional Snow Mapping: cost per task

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

  • Uncached input
  • Cached input
  • Output
  • Not split by token type
InSAR Regional Snow Mapping: cost per taskValues in US dollars: Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2: $0.0061; Claude Haiku 5.5 (xhigh), Claude Code: $0.0079; Nemotron 3.5 Super 120B (Reasoning), Terminus-2: $0.0081; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: $0.012; Mistral Large 4 (high), Terminus-2: $0.028; Laguna XS 2.1 (Reasoning), Terminus-2: $0.037; Muse Glimmer 30B (high), Terminus-2: $0.038; Qwen 3.8 27B (xhigh), Terminus-2: $0.049; Claude Sonnet 5.5 (xhigh), Claude Code: $0.055; Claude Haiku 4.5 (Reasoning), Claude Code: $0.068; Claude Sonnet 5 (high), Claude Code: $0.090; Step 5 Preview (Reasoning), Terminus-2: $0.093; Claude Opus 5.5 (xhigh), Claude Code: $0.14; Claude Opus 5 (high), Claude Code: $0.15; Claude Opus 4.8 (high), Claude Code: $0.15; Claude Opus 4.7 (high), Claude Code: $0.20$0$0.05$0.10$0.15$0.20Nemotron 3.5 Super VL 120B (Reasoning)Terminus-2$0.0061Claude Haiku 5.5 (xhigh)Claude Code$0.0079Nemotron 3.5 Super 120B (Reasoning)Terminus-2$0.0081Nemotron 3.5 Lightning 30B (Reasoning)Terminus-2$0.012Mistral Large 4 (high)Terminus-2$0.028Laguna XS 2.1 (Reasoning)Terminus-2$0.037Muse Glimmer 30B (high)Terminus-2$0.038Qwen 3.8 27B (xhigh)Terminus-2$0.049Claude Sonnet 5.5 (xhigh)Claude Code$0.055Claude Haiku 4.5 (Reasoning)Claude Code$0.068Claude Sonnet 5 (high)Claude Code$0.090Step 5 Preview (Reasoning)Terminus-2$0.093Claude Opus 5.5 (xhigh)Claude Code$0.14Claude Opus 5 (high)Claude Code$0.15Claude Opus 4.8 (high)Claude Code$0.15Claude Opus 4.7 (high)Claude Code$0.20
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: InSAR Regional Snow Mapping: cost per task
InSAR Regional Snow Mapping: cost per task
ModelDeveloperUncached inputCached inputOutputNot split by token typeTotal
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2NVIDIA$0.0036$0.00032$0.0021$0.00$0.0061
Claude Haiku 5.5 (xhigh), Claude CodeAnthropic$0.0016$0.0021$0.0024$0.0016$0.0079
Nemotron 3.5 Super 120B (Reasoning), Terminus-2NVIDIA$0.0049$0.00072$0.0025$0.000000050$0.0081
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2NVIDIA$0.0023$0.0070$0.0025$0.00000039$0.012
Mistral Large 4 (high), Terminus-2Mistral AI$0.011$0.0033$0.014$0.00000048$0.028
Laguna XS 2.1 (Reasoning), Terminus-2Poolside$0.0045$0.030$0.0031$0.00$0.037
Muse Glimmer 30B (high), Terminus-2Meta$0.023$0.0031$0.012$0.00000049$0.038
Qwen 3.8 27B (xhigh), Terminus-2Alibaba (Qwen)$0.020$0.0037$0.025$0.00$0.049
Claude Sonnet 5.5 (xhigh), Claude CodeAnthropic$0.017$0.0058$0.015$0.017$0.055
Claude Haiku 4.5 (Reasoning), Claude CodeAnthropic$0.011$0.022$0.023$0.011$0.068
Claude Sonnet 5 (high), Claude CodeAnthropic$0.020$0.028$0.021$0.020$0.090
Step 5 Preview (Reasoning), Terminus-2StepFun$0.044$0.0098$0.039$0.00$0.093
Claude Opus 5.5 (xhigh), Claude CodeAnthropic$0.045$0.015$0.040$0.045$0.14
Claude Opus 5 (high), Claude CodeAnthropic$0.039$0.031$0.036$0.039$0.15
Claude Opus 4.8 (high), Claude CodeAnthropic$0.045$0.026$0.035$0.045$0.15
Claude Opus 4.7 (high), Claude CodeAnthropic$0.051$0.058$0.036$0.050$0.20
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2NVIDIANo dataNo dataNo dataNo dataNo data

Open weights

InSAR Regional Snow Mapping: cost per task

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

  • Uncached input
  • Cached input
  • Output
  • Not split by token type
InSAR Regional Snow Mapping: cost per taskValues in US dollars: Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2: $0.0061; Nemotron 3.5 Super 120B (Reasoning), Terminus-2: $0.0081; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: $0.012; Mistral Large 4 (high), Terminus-2: $0.028; Laguna XS 2.1 (Reasoning), Terminus-2: $0.037; Muse Glimmer 30B (high), Terminus-2: $0.038; Qwen 3.8 27B (xhigh), Terminus-2: $0.049; Step 5 Preview (Reasoning), Terminus-2: $0.093$0$0.025$0.05$0.075$0.10Nemotron 3.5 Super VL 120B (Reasoning)Terminus-2$0.0061Nemotron 3.5 Super 120B (Reasoning)Terminus-2$0.0081Nemotron 3.5 Lightning 30B (Reasoning)Terminus-2$0.012Mistral Large 4 (high)Terminus-2$0.028Laguna XS 2.1 (Reasoning)Terminus-2$0.037Muse Glimmer 30B (high)Terminus-2$0.038Qwen 3.8 27B (xhigh)Terminus-2$0.049Step 5 Preview (Reasoning)Terminus-2$0.093
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: InSAR Regional Snow Mapping: cost per task
InSAR Regional Snow Mapping: cost per task
ModelDeveloperUncached inputCached inputOutputNot split by token typeTotal
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2NVIDIA$0.0036$0.00032$0.0021$0.00$0.0061
Nemotron 3.5 Super 120B (Reasoning), Terminus-2NVIDIA$0.0049$0.00072$0.0025$0.000000050$0.0081
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2NVIDIA$0.0023$0.0070$0.0025$0.00000039$0.012
Mistral Large 4 (high), Terminus-2Mistral AI$0.011$0.0033$0.014$0.00000048$0.028
Laguna XS 2.1 (Reasoning), Terminus-2Poolside$0.0045$0.030$0.0031$0.00$0.037
Muse Glimmer 30B (high), Terminus-2Meta$0.023$0.0031$0.012$0.00000049$0.038
Qwen 3.8 27B (xhigh), Terminus-2Alibaba (Qwen)$0.020$0.0037$0.025$0.00$0.049
Step 5 Preview (Reasoning), Terminus-2StepFun$0.044$0.0098$0.039$0.00$0.093
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2NVIDIANo dataNo dataNo dataNo dataNo data

Speed

All

InSAR Regional Snow Mapping: time per task

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

InSAR Regional Snow Mapping: time per taskInSAR Regional Snow Mapping: time per task: Claude Opus 4.7 (high), Claude Code: 0.3 min; Claude Opus 4.8 (high), Claude Code: 0.3 min; Claude Opus 5 (high), Claude Code: 0.4 min; Claude Sonnet 5.5 (xhigh), Claude Code: 0.4 min; Claude Sonnet 5 (high), Claude Code: 0.5 min; Claude Opus 5.5 (xhigh), Claude Code: 0.5 min; Claude Haiku 4.5 (Reasoning), Claude Code: 0.9 min; Claude Haiku 5.5 (xhigh), Claude Code: 1.0 min; Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2: 1.3 min; Nemotron 3.5 Super 120B (Reasoning), Terminus-2: 1.5 min; Mistral Large 4 (high), Terminus-2: 1.9 min; Muse Glimmer 30B (high), Terminus-2: 2.0 min; Qwen 3.8 27B (xhigh), Terminus-2: 2.4 min; Step 5 Preview (Reasoning), Terminus-2: 2.6 min; Laguna XS 2.1 (Reasoning), Terminus-2: 3.3 min; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 3.4 min; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 4.9 min0 min1.5 min3 min4.5 min6 minClaude Opus 4.7 (high)Claude Code0.3 minClaude Opus 4.8 (high)Claude Code0.3 minClaude Opus 5 (high)Claude Code0.4 minClaude Sonnet 5.5 (xhigh)Claude Code0.4 minClaude Sonnet 5 (high)Claude Code0.5 minClaude Opus 5.5 (xhigh)Claude Code0.5 minClaude Haiku 4.5 (Reasoning)Claude Code0.9 minClaude Haiku 5.5 (xhigh)Claude Code1.0 minNemotron 3.5 Super VL 120B (Reasoning)Terminus-21.3 minNemotron 3.5 Super 120B (Reasoning)Terminus-21.5 minMistral Large 4 (high)Terminus-21.9 minMuse Glimmer 30B (high)Terminus-22.0 minQwen 3.8 27B (xhigh)Terminus-22.4 minStep 5 Preview (Reasoning)Terminus-22.6 minLaguna XS 2.1 (Reasoning)Terminus-23.3 minNemotron 3.5 Lightning 30B (Reasoning)Terminus-23.4 minNemotron 3 Nano Omni 30B (Reasoning)Terminus-24.9 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: InSAR Regional Snow Mapping: time per task

Open weights

InSAR Regional Snow Mapping: time per task

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

InSAR Regional Snow Mapping: time per taskInSAR Regional Snow Mapping: time per task: Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2: 1.3 min; Nemotron 3.5 Super 120B (Reasoning), Terminus-2: 1.5 min; Mistral Large 4 (high), Terminus-2: 1.9 min; Muse Glimmer 30B (high), Terminus-2: 2.0 min; Qwen 3.8 27B (xhigh), Terminus-2: 2.4 min; Step 5 Preview (Reasoning), Terminus-2: 2.6 min; Laguna XS 2.1 (Reasoning), Terminus-2: 3.3 min; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 3.4 min; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 4.9 min0 min1.5 min3 min4.5 min6 minNemotron 3.5 Super VL 120B (Reasoning)Terminus-21.3 minNemotron 3.5 Super 120B (Reasoning)Terminus-21.5 minMistral Large 4 (high)Terminus-21.9 minMuse Glimmer 30B (high)Terminus-22.0 minQwen 3.8 27B (xhigh)Terminus-22.4 minStep 5 Preview (Reasoning)Terminus-22.6 minLaguna XS 2.1 (Reasoning)Terminus-23.3 minNemotron 3.5 Lightning 30B (Reasoning)Terminus-23.4 minNemotron 3 Nano Omni 30B (Reasoning)Terminus-24.9 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: InSAR Regional Snow Mapping: time per task

All

InSAR Regional Snow Mapping: score vs. time per task

InSAR Regional Snow Mapping score against time per task (at 100 output tokens/s), minutes · Higher and further left is better

  • Most attractive quadrant
  • Pareto line
InSAR Regional Snow Mapping: score vs. time per taskInSAR Regional Snow Mapping score (0 to 100) against time per task (minutes), log scale: Laguna XS 2.1 (Reasoning), Terminus-2: 87.5, 3.3 min; Mistral Large 4 (high), Terminus-2: 100, 1.9 min; Muse Glimmer 30B (high), Terminus-2: 95.83, 2.0 min; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 33.33, 4.9 min; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 79.17, 3.4 min; Nemotron 3.5 Super 120B (Reasoning), Terminus-2: 95.83, 1.5 min; Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2: 100, 1.3 min; Qwen 3.8 27B (xhigh), Terminus-2: 100, 2.4 min; Step 5 Preview (Reasoning), Terminus-2: 100, 2.6 min; Claude Haiku 4.5 (Reasoning), Claude Code: 95.83, 0.9 min; Claude Opus 5.5 (xhigh), Claude Code: 100, 0.5 min; Claude Sonnet 5.5 (xhigh), Claude Code: 100, 0.4 min; Claude Haiku 5.5 (xhigh), Claude Code: 100, 1.0 min; Claude Opus 5 (high), Claude Code: 100, 0.4 min; Claude Sonnet 5 (high), Claude Code: 100, 0.5 min; Claude Opus 4.8 (high), Claude Code: 100, 0.3 min; Claude Opus 4.7 (high), Claude Code: 100, 0.3 minMost attractive quadrant20304050607080901000.2 min0.5 min1 min2 min5 min10 minTime per task (minutes, log scale)InSAR Regional Snow Mapping scoreNemotron 3.5 Super VL 120B (Reasoning), Terminus-2Claude Opus 5.5 (xhigh), Claude CodeClaude Sonnet 5.5 (xhigh), Claude CodeClaude Opus 4.7 (high), Claude CodeMistral Large 4 (high), Terminus-2Qwen 3.8 27B (xhigh), Terminus-2Step 5 Preview (Reasoning), Terminus-2Claude Haiku 5.5 (xhigh), Claude CodeClaude Opus 5 (high), Claude CodeClaude Sonnet 5 (high), Claude CodeClaude Opus 4.8 (high), Claude CodeMuse Glimmer 30B (high), Terminus-2Nemotron 3.5 Super 120B (Reasoning), Terminus-2Claude Haiku 4.5 (Reasoning), Claude CodeLaguna XS 2.1 (Reasoning), Terminus-2Nemotron 3.5 Lightning 30B (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 InSAR Regional Snow Mapping 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: InSAR Regional Snow Mapping: score vs. time per task

Open weights

InSAR Regional Snow Mapping: score vs. time per task

InSAR Regional Snow Mapping score against time per task (at 100 output tokens/s), minutes · Higher and further left is better

  • Most attractive quadrant
  • Pareto line
InSAR Regional Snow Mapping: score vs. time per taskInSAR Regional Snow Mapping score (0 to 100) against time per task (minutes), log scale: Laguna XS 2.1 (Reasoning), Terminus-2: 87.5, 3.3 min; Mistral Large 4 (high), Terminus-2: 100, 1.9 min; Muse Glimmer 30B (high), Terminus-2: 95.83, 2.0 min; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 33.33, 4.9 min; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 79.17, 3.4 min; Nemotron 3.5 Super 120B (Reasoning), Terminus-2: 95.83, 1.5 min; Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2: 100, 1.3 min; Qwen 3.8 27B (xhigh), Terminus-2: 100, 2.4 min; Step 5 Preview (Reasoning), Terminus-2: 100, 2.6 minMost attractive quadrant20304050607080901001 min2 min5 min10 minTime per task (minutes, log scale)InSAR Regional Snow Mapping scoreMistral Large 4 (high), Terminus-2Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2Qwen 3.8 27B (xhigh), Terminus-2Step 5 Preview (Reasoning), Terminus-2Muse Glimmer 30B (high), Terminus-2Nemotron 3.5 Super 120B (Reasoning), Terminus-2Laguna XS 2.1 (Reasoning), Terminus-2Nemotron 3.5 Lightning 30B (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 InSAR Regional Snow Mapping 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: InSAR Regional Snow Mapping: score vs. time per task

Score vs. release date

All

InSAR Regional Snow Mapping: score vs. release date

InSAR Regional Snow Mapping score against the model's public release date

InSAR Regional Snow Mapping: score vs. release dateInSAR Regional Snow Mapping score (0 to 100) against release date: Laguna XS 2.1 (Reasoning), Terminus-2: 87.5, 2 July 2026; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 33.33, 28 April 2026; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 79.17, 11 August 2026; Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2: 100, 15 October 2026; Qwen 3.8 27B (xhigh), Terminus-2: 100, 14 August 2026; Claude Haiku 4.5 (Reasoning), Claude Code: 95.83, 15 October 2025; Claude Opus 5.5 (xhigh), Claude Code: 100, 22 September 2026; Claude Sonnet 5.5 (xhigh), Claude Code: 100, 28 September 2026; Claude Haiku 5.5 (xhigh), Claude Code: 100, 7 October 2026; Claude Opus 5 (high), Claude Code: 100, 24 July 2026; Claude Sonnet 5 (high), Claude Code: 100, 30 June 2026; Claude Opus 4.8 (high), Claude Code: 100, 28 May 2026; Claude Opus 4.7 (high), Claude Code: 100, 16 April 20262030405060708090100Oct 2025Jan 2026Apr 2026Jul 2026Oct 2026Jan 2027Release dateInSAR Regional Snow Mapping scoreNemotron 3.5 Super VL 120B (Reasoning), Terminus-2Qwen 3.8 27B (xhigh), Terminus-2Claude Opus 5.5 (xhigh), Claude CodeClaude Sonnet 5.5 (xhigh), Claude CodeClaude Haiku 5.5 (xhigh), Claude CodeClaude Opus 5 (high), Claude CodeClaude Sonnet 5 (high), Claude CodeClaude Opus 4.8 (high), Claude CodeClaude Opus 4.7 (high), Claude CodeClaude Haiku 4.5 (Reasoning), Claude CodeLaguna XS 2.1 (Reasoning), Terminus-2Nemotron 3.5 Lightning 30B (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 InSAR Regional Snow Mapping score. Models with no published release date are not plotted.

Data table: InSAR Regional Snow Mapping: score vs. release date

Open weights

InSAR Regional Snow Mapping: score vs. release date

InSAR Regional Snow Mapping score against the model's public release date

InSAR Regional Snow Mapping: score vs. release dateInSAR Regional Snow Mapping score (0 to 100) against release date: Laguna XS 2.1 (Reasoning), Terminus-2: 87.5, 2 July 2026; Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2: 33.33, 28 April 2026; Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2: 79.17, 11 August 2026; Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2: 100, 15 October 2026; Qwen 3.8 27B (xhigh), Terminus-2: 100, 14 August 20262030405060708090100Mar 2026May 2026Jul 2026Sep 2026Nov 2026Release dateInSAR Regional Snow Mapping scoreNemotron 3.5 Super VL 120B (Reasoning), Terminus-2Qwen 3.8 27B (xhigh), Terminus-2Laguna XS 2.1 (Reasoning), Terminus-2Nemotron 3.5 Lightning 30B (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 InSAR Regional Snow Mapping score. Models with no published release date are not plotted.

Data table: InSAR Regional Snow Mapping: score vs. release date

Background

Our first tasks on synthetic aperture radar (SAR): 64 Sentinel-1 interferogram windows from January 2021 to March 2023 over the Trentino Alps, with phase, coherence and displacement layers, two weather layers and modelled snow depth. The source has no agent questions, so we wrote 24: phase and coherence checks, lining the weather layers up with the radar, and regional statistics.

Licence: dataset card MIT; the Copernicus Sentinel-1, IT-SNOW (ISPRA) and ERA5 (ECMWF, C3S) terms apply to their layers.

Published by the LINKS Foundation (Barco and colleagues).

Related links

Explore evaluations