# Agenticflow LLM Models > List, filter, and recommend LLM models available in the AgenticFlow AI workspace for use in agents and workforce nodes. Should trigger whenever the user mentions LLM models, choosing a provider/model for an agent, model capabilities (reasoning, speed, cost), model selection for specific tasks, or understanding which models are available and their trade-offs. Use the live `af get /models` as the authoritative source. The rest of this skill provides recommendations and context but never overrides the live response. - Skill: `antongulin/agenticflow-llm-models` (Agent Skill) - Install (CLI): `npx skillmds@latest add antongulin/agenticflow-llm-models` - Raw SKILL.md: https://api.skillmd.com/api/skills/antongulin/agenticflow-llm-models/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: antongulin (https://skillmd.com/u/antongulin) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/antongulin/agenticflow-llm-models --- > **Author**: Anton Gulin · **Tool**: [opencode-skill-creator](https://github.com/antongulin/opencode-skill-creator) · **GitHub**: [@antongulin](https://github.com/antongulin) · **Registry**: [skills.sh](https://www.skills.sh/docs) # AgenticFlow LLM Models Choose the right model for your agent based on capability needs, speed requirements, and reasoning depth. Use built-in credits for all models listed here. ## When NOT to use this skill Use `agenticflow-built-in-credits` skill instead for pricing, credits, or billing questions. Use `agenticflow-mcp` skill if they need external API keys (BYOK). This skill covers model selection and capabilities, not account management or credit usage. ## Orient first ```bash af bootstrap --json ``` Extract `models[]` — this is the **source of truth** for available models in your workspace. Never hardcode model lists; they change between CLI releases and backend deployments. The models below are recommendations, but the live `models[]` array is the final authority. ## Discover & health ```bash af changelog --json # What's new in the CLI — model additions/removals af context --json # AI agent orientation, env vars, invocation guidance af bootstrap --strict --json # Health check — exits non-zero if degraded ``` `af bootstrap` returns an `invocation` block and `data_fresh` boolean. If `data_fresh: false`, the backend is degraded — don't rely on stale model data from a degraded response. `af bootstrap --strict` exits non-zero when the backend is unhealthy, so CI/automation can abort before choosing models against a degraded workspace. > **Verification rule:** Before recommending any model, check `models[]` from `af bootstrap --json`. If a model is absent from that list, warn the user and fall back to a confirmed model. --- ## Author's Top 3 Recommendations These are the author's personal picks based on reliability, reasoning quality, and speed: | Rank | Model | Role | Why | |------|-------|------|-----| | 1st | `deepseek-v4-flash` | **Primary default** | Best all-rounder — strong reasoning, reliable tool use, good speed. Replaces the older GLM 4.7 Flash default. | | 2nd | `gemini-2.5-flash-lite` | **Fallback / media** | Fastest option with media support. Use when speed matters more than reasoning depth. | | 3rd | `qwen-3.5-flash` | **Deep verification** | Deepest thinker — use when reasoning depth and verification matter most. | > **Note:** `deepseek-v4-flash` and `gemini-2.5-flash-lite` may not appear in every CLI release's hardcoded `KNOWN_MODELS` list if they ship between releases. Always verify against `af bootstrap --json > models[]`. If absent, the backend may still serve them — proceed with a dry-run to confirm. --- ## Upstream Canonical Models The CLI's built-in validator recognizes these models as of v1.10.5. They are always safe to use: ``` agenticflow/deepseek-v3.2 agenticflow/gemma-4-31b-it agenticflow/gemma-4-26b-a4b-it agenticflow/gemini-2.0-flash agenticflow/gpt-4o-mini agenticflow/qwen-3.5-flash ``` ### Model characteristics | Model | Speed | Reasoning | Best for | |-------|-------|-----------|----------| | `agenticflow/deepseek-v3.2` | Medium | Strong | Reliable reasoning, good tool use | | `agenticflow/gemma-4-31b-it` | Fast | Light | General purpose, high-volume | | `agenticflow/gemma-4-26b-a4b-it` | Fast | Light | General purpose, slightly smaller context | | `agenticflow/gemini-2.0-flash` | Fast | Light | **Deprecated** — still served but being replaced by 2.5 Flash Lite | | `agenticflow/gpt-4o-mini` | Fast | Light | **Default for blueprints** (v1.8.1+) — follows system prompts reliably, good for tool calling | | `agenticflow/qwen-3.5-flash` | Medium-Deep | Very strong | Deep verification, complex reasoning | ### Default model change (v1.8.1+) - **Before v1.8.1:** Default was `agenticflow/gemini-2.0-flash` - **After v1.8.1:** Default is `agenticflow/gpt-4o-mini` - **Reason for change:** Gemini 2.0 Flash refuses `web_search` on "latest X" prompts citing knowledge cutoff, even with explicit system prompt rules. GPT-4o-mini follows system prompts and calls tools reliably. --- ## Model selection guide **Need a default?** ```bash # Author's primary recommendation — deepseek-v4-flash af agent create --body '{"name":"My Agent","model":"deepseek-v4-flash","project_id":""}' --json # Or use the upstream blueprint default — gpt-4o-mini af agent create --body '{"name":"My Agent","model":"agenticflow/gpt-4o-mini","project_id":""}' --json # For maximum speed with media support af agent create --body '{"name":"My Agent","model":"gemini-2.5-flash-lite","project_id":""}' --json ``` **Need reasoning?** - Deep verification: `qwen-3.5-flash` — deepest thinker - Reliable tool use + reasoning: `deepseek-v3.2` or `deepseek-v4-flash` **Need speed only?** - Fastest correct: `gpt-4o-mini` or `gemma-4-31b-it` --- ## Workforce model selection All agents in a workforce inherit the model: ```bash # Default (v1.8.1+) af workforce init --blueprint dev-shop --model agenticflow/gpt-4o-mini --name "My Team" --json # Or use author's primary pick af workforce init --blueprint dev-shop --model deepseek-v4-flash --name "My Team" --json ``` --- ## Reasoning configuration Expose reasoning tokens (where supported): ```bash af schema agent --field model_user_config --json ``` Check the full agent schema (for all fields): ```bash af schema agent --json ``` For models with hidden reasoning (e.g. some Gemini variants), configure via `thinking_config` to expose reasoning tokens. --- ## Verify model availability ```bash # Always dry-run first af agent create --body @agent.json --dry-run --json ``` The CLI validates the model string at create time. Typos fail fast with an actionable hint listing known models. If you pass a `vendor/model-name`-shaped string not in the known list, the CLI warns but allows it to proceed — so brand-new models work before the CLI is updated. --- ## Avoid these models Based on upstream changelog and known issues: | Model | Issue | |-------|-------| | `agenticflow/gemini-2.0-flash` | Deprecated, replaced by 2.5 Flash Lite. Still served but default changed to gpt-4o-mini. | | `agenticflow/gemini-2.0-flash-lite` | Deprecated, replaced by 2.5 Flash Lite | > In general, if a model is absent from `af bootstrap --json > models[]`, it may have been deprecated or renamed. Check the `hint` field on 400/422 errors for alternatives. --- ## Fallback model guide If your preferred model is unavailable: 1. Run `af bootstrap --json` and check `models[]` 2. Pick the closest match from the confirmed list above 3. Use `--dry-run` on create to validate before deploying 4. For reasoning-heavy tasks: fall back to `qwen-3.5-flash` or `deepseek-v3.2` 5. For speed-first tasks: fall back to `gpt-4o-mini` or `gemma-4-31b-it` --- ## Cleanup Test agents consume credits. Delete when done: ```bash af agent delete --agent-id --json ``` --- ## On errors - **400 / Invalid model** → Check `models[]` from bootstrap; model may have been renamed or is not yet in the CLI's hardcoded list. Try `--dry-run` first. - **402 / Payment Required** → Model requires credits; see `agenticflow-built-in-credits` skill - **422 / Model not available** → Model temporarily unavailable; the `hint` suggests alternatives - **finish_reason=length** → Increase `max_tokens` in `model_user_config` When `hint` is non-empty, follow it before retrying.