Best for coding
—
—
Highest coding score in the current SENA model index.
SENA Model IndexIndependent AI model evaluationUpdated regularly
SENA Models
Which AI model is actually better?
Start with SENA's neutral benchmark, then configure what matters to your workload. Weight coding, reasoning, agents, reliability, speed and cost to find the model that best fits how you actually plan to use AI.
SENA testing philosophy
We test what models can actually do — not just what benchmark scores say they can do.
Read our methodologyStart with the job
What do you need the model to do?
Pick a workload and SENA will re-weight the benchmark for that job. You can fine-tune every weight and constraint afterwards.
Configure your benchmark
What are you actually choosing a model for?
The SENA score remains the neutral benchmark. Your Fit score re-weights the same model data around your workload, then applies practical constraints such as context window and input price.
Importance weights
0 = ignore · 10 = critical
Coding5/10
Reasoning5/10
Agents5/10
Documents4/10
Extraction3/10
Long context3/10
Reliability5/10
Speed2/10
Value3/10
Hard requirements
Models that fail these are excluded.
Recommended for your configuration
No models match every requirement
Relax the context, price or verification requirement to see a recommendation.
SENA Model Index
Best models for your workload
Higher score is better
Loading SENA model index…
Quick answers
The best model depends on the job.
Best for agents
—
—
Highest agent-workflow score in the current SENA model index.
Best for documents
—
—
Highest document score in the current SENA model index.
Best value
—
—
Highest cost-adjusted value score in the current SENA model index.
Side by side
Compare AI models
| Metric |
|---|
| Your Fit score |
| SENA score |
| Coding |
| Reasoning |
| Agents |
| Documents |
| Extraction |
| Long context |
| Reliability |
| Speed |
| Value |
| Context window |
| Max output |
| Input price |
| Output price |
| Median latency |
Cost-adjusted performance
The best model isn't always the most expensive one.
SENA tracks capability alongside provider token pricing so teams can compare model quality against the economics of deploying each model. Verified runs add measured cost and latency when available.
Deployment economics
What will this model actually cost you?
Enter a rough production workload. SENA converts provider token prices into an estimated monthly API bill so you can compare capability and economics in the same decision.
Estimate uses the input/output token prices currently supplied with each benchmark model. Cached-input discounts and other provider-specific pricing tiers are not included.
Model
Fit
Per 1K req.
Est. / month
Pricing is not available for the current benchmark rows.
Provider model catalog
A provider can have dozens of models.
The catalog answers “what can I use?” while the SENA benchmark answers “how do comparable models perform?”. Image, realtime, speech, embedding and safety models stay visible without being forced into an apples-to-oranges leaderboard.
OpenAI catalog reference
50 current entries seeded
Kept separate from SENA benchmark coverage so the page can scale as providers add model families.
50 models shown0 currently benchmarked by SENA
OpenAI/Frontier
GPT-5.6 Sol
GPT-5.6
Frontier professional-work model.
OpenAI/Frontier
GPT-5.6 Terra
GPT-5.6
Balanced intelligence and cost.
OpenAI/Frontier
GPT-5.6 Luna
GPT-5.6
Cost-sensitive GPT-5.6 option.
OpenAI/General
GPT-5.4
GPT-5.4
GPT-5.4 model in the general catalog.
OpenAI/General
GPT-5.4 Pro
GPT-5.4
GPT-5.4 model in the general catalog.
OpenAI/General
GPT-5.4 mini
GPT-5.4
GPT-5.4 model in the general catalog.
OpenAI/General
GPT-5.4 nano
GPT-5.4
GPT-5.4 model in the general catalog.
OpenAI/General
GPT-5.5
GPT-5.5
GPT-5.5 model in the general catalog.
OpenAI/General
GPT-5.5 Pro
GPT-5.5
GPT-5.5 model in the general catalog.
OpenAI/Coding
GPT-5.3-Codex
Codex
Codex model in the coding catalog.
OpenAI/General
GPT-5.2
GPT-5
GPT-5 model in the general catalog.
OpenAI/General
GPT-5.2 Pro
GPT-5
GPT-5 model in the general catalog.
OpenAI/General
GPT-5.1
GPT-5
GPT-5 model in the general catalog.
OpenAI/General
GPT-5
GPT-5
GPT-5 model in the general catalog.
OpenAI/General
GPT-5 mini
GPT-5
GPT-5 model in the general catalog.
OpenAI/General
GPT-5 nano
GPT-5
GPT-5 model in the general catalog.
OpenAI/General
GPT-5 Pro
GPT-5
GPT-5 model in the general catalog.
OpenAI/General
o3-pro
o-series
o-series model in the general catalog.
Methodology
How SENA evaluates AI models
01
Capability dimensions
SENA compares coding, reasoning, agents, documents, extraction and long-context capability separately.
02
Editorial baseline
Editorial rows are provisional SENA assessments used to make the index useful before a verified live run is available.
03
Official model economics
Context windows, output limits and token pricing are surfaced from provider documentation and kept separate from capability scores.
04
Clear verification status
Every model is labelled Editorial or Verified so readers can distinguish curated assessments from executed benchmark results.
05
Repeatable live tests
When verification is enabled, SENA runs the same task set across providers and stores the resulting scores, cost and latency.
06
Cost-aware ranking
Value is considered alongside capability so teams can compare strong models without ignoring deployment economics.
Task templates
—
Live model calls
—
Live runs / task
—
Data updated
—
How to choose the right AI model
There is no universally best large language model. The right model depends on the workload, required accuracy, latency tolerance, context size, tool requirements and inference budget.
Coding-heavy applications may benefit from models that perform strongly on repository reasoning and tool use, while document-processing workloads may prioritize long context, retrieval accuracy and lower token cost.
SENA evaluates these dimensions separately so teams can compare models based on the work they actually need to perform rather than relying on a single generalized benchmark score.
SENA Intelligence