Which AI model should I use today?
Checked Oct 12, 2026
Sonnet 5.5Pick
$2 / $10
Anthropic's best model for agentic coding (70.6% on Terminal-Bench 4.0), at half the price of Opus.
model: "claude-sonnet-5-5", output_config: { effort: "high" }Opus 5.5
$4 / $20
Twice the price and lower on agentic coding (66.4%). Use it to plan the architecture, then hand the coding to Sonnet.
Fable 5.1
$10 / $50
Five times the price. Keep it for a bug that Sonnet and Opus both failed to fix.
Haiku 5.5
$0.10 / $0.50
A cheap subagent for side jobs. Keep a stronger model as your main coder.
Sonnet 5Older
$2 / $10
The previous generation, at the same price, so there's no reason to stay on it.
Medium effort is enough for small, well-scoped edits.
Prices are per million tokens, input / output. We pick for the best result per dollar, so a pricier model can score higher and still lose out. This pick last changed Oct 9, 2026. Sources: Anthropic docs, OpenAI docs, Gemini docs, Artificial Analysis, Vals AI.
Maintained by the AI team at Vegan Hacktivists
| Lab | Task | Model | Effort | Why |
|---|---|---|---|---|
| Claude | Code | Sonnet 5.5 claude-sonnet-5-5 | High | Anthropic's best model for agentic coding (70.6% on Terminal-Bench 4.0), at half the price of Opus. |
| Claude | Everyday | Sonnet 5.5 claude-sonnet-5-5 | Low | Handles emails, drafts and quick questions well, and low effort keeps it fast and cheap. |
| Claude | Research | Opus 5.5 claude-opus-5-5 | High | Top score on the Artificial Analysis Intelligence Index (58, measured at max effort) at 40% of Fable's price. |
| Claude | Agents | Sonnet 5.5 claude-sonnet-5-5 | Medium | Ties Opus on the Vals Index (67.0%, measured at max effort) at about two-thirds of the cost per task. |
| OpenAI | Code | GPT-6.1 Sol gpt-6.1-sol | High | Beats Astra's best on DeepSWE (75.2% vs 74.1%) at a fraction of the cost per task. |
| OpenAI | Everyday | GPT-6 Luna gpt-6-luna | Low | Handles emails, drafts and quick questions at 20x less than 6.1 Sol, and low effort keeps it quick. |
| OpenAI | Research | GPT-6 Astra gpt-6-astra | High | Leads on hard math and science (96.1% on GPQA Diamond). OpenAI's strongest model for deep work. |
| OpenAI | Agents | GPT-6.1 Sol gpt-6.1-sol | High | Edges Claude Opus 5.5 on AutomationBench at about a third of the cost. |
| Gemini | Code | 3.8 Flash gemini-3.8-flash | Medium | The strongest Google model for code that you can use today. Beats 3.1 Pro on coding tests at about a third of the price. |
| Gemini | Everyday | 3.8 Flash gemini-3.8-flash | Low | Google recommends it for drafts and chat, and it's fast and cheap. |
| Gemini | Research | 3.8 Flash gemini-3.8-flash | High | Beats 3.1 Pro on 9 of 13 shared benchmarks at about a third of the price. |
| Gemini | Agents | 3.8 Flash gemini-3.8-flash | Medium | Powers Google's own managed agents. Medium is Google's recommended level for agent work. |
Want helpputting this to work?
Vegan Hacktivists offers free AI and automation support to animal protection organizations — from a first conversation to a full build.
