Best overall
Closed model
Anthropic
- Intelligence
- 58.4
- Speed
- 96.6tok/s
- Input / output
- $4 / $20
- Context
- 1M
Best fit
Demanding reasoning, coding and long-horizon agents
AI · LEADERBOARD
Una comparación práctica de los modelos de IA que los equipos pueden desplegar hoy.
Independent editorial snapshot · not live platform telemetry
Dated editorial snapshot
Review target: weekly and after major releases. Rankings reflect this dated snapshot. OpenRouter usage below has its own source timestamp.
The short answer
Best overall
Closed model
Anthropic
Best fit
Demanding reasoning, coding and long-horizon agents
Best open model
Open weights
DeepSeek
Best fit
Text and image workloads with a sparse MoE model
Best value
Closed model
Anthropic
Best fit
High-volume summarization, subagents and browser use
OpenRouter usage
Top models by prompt, completion, and reasoning tokens processed on OpenRouter. Showing the first 20 only.
From benchmark to decision
Nine recurring production scenarios, with the trade-offs that matter before a team commits to a model.
Selected use case
01 / 09
Decision view
Why it fits
How the ranking works
The ranking combines public benchmark signals with price, measured throughput, context and deployment constraints. It is a decision aid, not a universal definition of model quality.
Value score
(intelligence − 25)² × min(1, speed ÷ 30) ÷ (input cost + output cost × 3)
Output is weighted ×3 to reflect output-heavy production use; intelligence is squared so cheap but weak models do not dominate.
Public references
Prices and performance can change without notice. Validate the finalists against your own prompts, infrastructure and risk requirements before procurement.
FAQ
The first row and Best overall card show the current intelligence-first leader. The best choice can change once price, latency, hosting and task-specific evaluation enter the decision.
The Best open model card is recalculated from models with verified downloadable weights, using the same intelligence-first ordering as the main table.
The Best value card is recalculated from the published value formula whenever verified intelligence, speed or base token pricing changes.
No. Context capacity only describes how much can be submitted. Retrieval design, instruction quality, attention behaviour and evaluation still determine whether that context produces a useful answer.
A scheduled pipeline revalidates the dataset weekly and after major releases. Failed or conflicting source checks keep the last verified snapshot online instead of publishing partial data.
The default order uses the public intelligence signal. Speed, price, context, openness and a separate value formula remain visible so teams can challenge the default order.
Tenten does not sell model tokens through this page. Public sources are linked, assumptions are shown, and we recommend a task-specific evaluation before any production decision.
TENTEN AI · MODEL EVALUATION
The leaderboard is the start
A model only becomes a good decision after it survives your documents, permissions, latency budget and failure cases.
Model explorer
No single score tells the whole story. Narrow the list, inspect the trade-offs, then compare the finalists side by side.
23models
| Compare models | Model | Best for | ||||||
|---|---|---|---|---|---|---|---|---|
| #1 | 58.4 | 96.6 tok/s | 17.4 | $4 / $20input / output | 1M | Demanding reasoning, coding and long-horizon agents | ||
| #2 | 56.9 | 78.2 tok/s | 6.3 | $10 / $50input / output | 1.1M | Advanced analysis, software engineering and research | ||
| #3 | 54.5 | 79.3 tok/s | 27.2 | $2 / $10input / output | 1.1M | Agentic coding, computer use and document-heavy work | ||
| #4 | 54.2 | 139.3 tok/s | 26.7 | $2 / $10input / output | 1M | Everyday feature development and bug fixes | ||
| #5 | 51.0 | 97.2 tok/s | 42.2 | $1 / $5input / output | 1.1M | Everyday agents, coding and screenshot-based work | ||
| #6 | 49.0 | 68.7 tok/s | 3.6 | $10 / $50input / output | 1M | Long code refactors and sustained agent workflows | ||
| #7 | 46.0 | 30.9 tok/s | 13.8 | $2 / $10input / output | 1.0M | Long-horizon coding with image and video input | ||
| #8 | 43.3 | 143.1 tok/s | 5.2 | $4 / $20input / output | 1.0M | Personal and professional agents with multimodal input | ||
| #9 | 36.2 | 235.2 tok/s | 78.9 | $0.10 / $0.50input / output | 1M | High-volume summarization, subagents and browser use | ||
| #10 | 34.3 | 86.8 tok/s | 4.4 | $2 / $6input / output | 500K | Long-running software engineering and knowledge work | ||
| #11 | 33.6 | 40.1 tok/s | 3.7 | $2 / $6input / output | 1M | Multimodal work with text, images and video | ||
| #12 | 32.5 | 175 tok/s | 28.8 | $0.15 / $0.60input / output | 1.0M | Text and image workloads with a sparse MoE model | ||
| #13 | 31.7 | 75.8 tok/s | 4.1 | $0.08 / $3.6input / output | 1.0M | Complex software engineering and long-horizon agents | ||
| #14 | 30.5 | 120.3 tok/s | 2.3 | $0.83 / $4.13input / output | 1.0M | Software engineering, agents and multi-step reasoning | ||
| #15 | 29.9 | 49.4 tok/s | 7.9 | $0.44 / $0.87input / output | 1.1M | Demanding multimodal work across text, images, video and audio | ||
| #16 | 28.9 | 118.9 tok/s | 15.7 | $0.08 / $0.30input / output | 1.0M | Complex agents, coding and cybersecurity workflows | ||
| #17 | 28.5 | 165.4 tok/s | 1.4 | $1 / $2.7input / output | 1.0M | Agentic software engineering with image and video input | ||
| #18 | 27.8 | 118.3 tok/s | 4.7 | $0.10 / $0.50input / output | 1.1M | High-volume chat, classification and lightweight agents | ||
| #19 | 26.0 | 161.3 tok/s | 0.2 | $0.70 / $2.1input / output | 1.0M | Multimodal reasoning, coding and agentic workloads | ||
| #20 | 24.7 | 139.3 tok/s | 0.0 | $1.25 / $4.25input / output | 1.0M | Agentic coding and repository-level tool use | ||
| #21 | 20.1 | 38.3 tok/s | 4.1 | $0.15 / $1.88input / output | 1M | Coding, research and long-running multimodal agents | ||
| #22 | 17.7 | 299.3 tok/s | 16.3 | $0.25 / $1input / output | 1.0M | Coding and long-horizon agentic workflows | ||
| #23 | 12.5 | 195.7 tok/s | 11.9 | $1 / $4.05input / output | 524K | General reasoning, coding and tool-using agents |
Best fit
Demanding reasoning, coding and long-horizon agents
Best fit
Advanced analysis, software engineering and research
Best fit
Agentic coding, computer use and document-heavy work
Best fit
Everyday feature development and bug fixes
Best fit
Everyday agents, coding and screenshot-based work
Best fit
Long code refactors and sustained agent workflows
Best fit
Long-horizon coding with image and video input
Best fit
Personal and professional agents with multimodal input
Best fit
High-volume summarization, subagents and browser use
Best fit
Long-running software engineering and knowledge work
Best fit
Multimodal work with text, images and video
Best fit
Text and image workloads with a sparse MoE model
Best fit
Complex software engineering and long-horizon agents
Best fit
Software engineering, agents and multi-step reasoning
Best fit
Demanding multimodal work across text, images, video and audio
Best fit
Complex agents, coding and cybersecurity workflows
Best fit
Agentic software engineering with image and video input
Best fit
High-volume chat, classification and lightweight agents
Best fit
Multimodal reasoning, coding and agentic workloads
Best fit
Agentic coding and repository-level tool use
Best fit
Coding, research and long-running multimodal agents
Best fit
Coding and long-horizon agentic workflows
Best fit
General reasoning, coding and tool-using agents