MODEL COMPARISONS
Compare AI models
Head-to-head pricing, context and Artificial Analysis benchmarks for the models people are weighing up right now. Run any of them with one SandBase API key.
Popular comparisons
The better value on each line is highlighted. Prices are per 1M tokens on SandBase.
The two frontier flagships of October 2026, tested head to head on the same prompts.
- 59.1
- Intelligence Index
- 50.4
- $10.04
- Cost per task
- $4.97
- 97 tok/s
- Output speed
- 86 tok/s
- $3.40
- Input / 1M tokens
- $4
Argon is rated level with GPT-6 Astra on intelligence at a lower cost per task.
- 56.9
- Intelligence Index
- 50.4
- $6.24
- Cost per task
- $4.97
- 44 tok/s
- Output speed
- 86 tok/s
- $9
- Input / 1M tokens
- $4
The default picks for long agentic coding sessions.
- 59.1
- Intelligence Index
- 56.9
- $10.04
- Cost per task
- $6.24
- 97 tok/s
- Output speed
- 44 tok/s
- $3.40
- Input / 1M tokens
- $9
When the flagship is worth it, and when Sonnet does the job for less.
- 59.1
- Intelligence Index
- 54.5
- $10.04
- Cost per task
- $10.02
- 97 tok/s
- Output speed
- 131 tok/s
- $3.40
- Input / 1M tokens
- $1.70
The low-cost frontier option from each lab.
- 54.6
- Intelligence Index
- 50.4
- $1.27
- Cost per task
- $4.97
- 64 tok/s
- Output speed
- 86 tok/s
- $2
- Input / 1M tokens
- $4
Models covered
Every model in a comparison above, with the numbers each page compares.
| Model | Input / 1M | Output / 1M | Context | IntelligenceAA | Compared with |
|---|---|---|---|---|---|
Claude Opus 5.5anthropic/claude-opus-5.5 | $3.40 | $17 | 1M | 59.1 | |
Gemini 4 Argongoogle/gemini-4-argon | $4 | $20 | 1M | 50.4 | |
GPT-6 Astraopenai/gpt-6-astra | $9 | $45 | 1.1M | 56.9 | |
Claude Sonnet 5.5anthropic/claude-sonnet-5.5 | $1.70 | $8.50 | 1M | 54.5 | |
GPT-6.1 Solopenai/gpt-6.1-sol | $2 | $10 | 1.1M | 54.6 |
Intelligence Index: Artificial Analysis, snapshot 2026-11-13. Full rankings: SandBase Lab leaderboards.
More resources
Free Tools
Token counter, LLM cost calculator, JSON, cURL, image and QR tools that run in your browser.
GUIDEAgent ecosystem
Runtimes, sandboxes, memory, tool protocols, gateways and observability projects for production Agents.
TOOLSandBase CLI
Connect Codex, Claude Code, Cursor and 20+ other AI clients to SandBase with one command.
PROGRAMHackathon support
API credits, sandboxed runtime, starter workflows and builder help for AI Agent hackathon teams.
CODEOpen source
SDKs, the CLI and working examples you can read, fork and run.
DIRECTORYMCP servers
Inspect what each MCP server exposes before attaching it to an Agent.

