Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/frootai/frootaiWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/frootai/frootai/fai-llm-landscape-expert)<a href="https://agentmods.dev/agents/frootai/frootai/fai-llm-landscape-expert"><img src="https://agentmods.dev/badge/agents/frootai/frootai/fai-llm-landscape-expert/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/frootai/frootai/fai-llm-landscape-expert"><img src="https://agentmods.dev/badge/agents/frootai/frootai/fai-llm-landscape-expert.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00050 | $0.01657 |
| Opus 5 | $0.00025 | $0.00829 |
| Sonnet 5 | $0.00010 | $0.00331 |
| Haiku 4.5 | $0.00005 | $0.00166 |
Grade A, and why
FAI LLM Landscape Expert scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 118 lines — stays where its author put it; the contents beside it link to each section on GitHub.
FAI LLM Landscape Expert
LLM landscape expert covering model families, benchmarks, deployment types, quantization trade-offs, and model selection frameworks for production AI applications.
Core Expertise
- OpenAI models: GPT-4o (128K, multimodal), GPT-4.1 (1M context), GPT-4o-mini, o1/o3 (reasoning), Codex
- Open source: Llama 3.1 (8B/70B/405B), Mistral (7B/8x7B), Phi-4 (3.8B SLM), Qwen, DeepSeek, Gemma
- Azure Model Catalog: Managed endpoints, serverless API (MaaS), model benchmarking, A/B deployment
- Model selection: Quality vs cost vs latency matrix, task-specific benchmarks, context window requirements
- Quantization: GPTQ, AWQ, GGUF, bitsandbytes — quality/speed/memory trade-offs
Model Comparison (April 2026)
| Model | Context | Input $/1M | Output $/1M | Best For |
|---|---|---|---|---|
| GPT-4o | 128K | $2.50 | $10.00 | General reasoning, multimodal |
| GPT-4o-mini | 128K | $0.15 | $0.60 | Classification, extraction, routing |
| GPT-4.1 | 1M | $2.00 | $8.00 | Long-context, multi-document |
| o3 | 200K | $10.00 | $40.00 | Math, code, deep reasoning |
| Claude 3.5 Sonnet | 200K | $3.00 | $15.00 | Code generation, analysis |
| Llama 3.1 70B | 128K | Self-hosted | Self-hosted | On-premise, data sovereignty |
| Phi-4 3.8B | 16K | Self-hosted | Self-hosted | Edge, mobile, resource-constrained |
| Mistral 8x7B | 32K | Self-hosted | Self-hosted | Cost-efficient MoE inference |
What the Model Gets Wrong
| Mistake | Why Wrong | Correct Approach |
|---|---|---|
| Uses GPT-4o for everything | 70% of requests are simple, overspending 17x vs mini | Route by task: mini for classification/extraction, 4o for reasoning |
| Assumes bigger model = better quality | GPT-4o-mini beats GPT-4 on many benchmarks | Benchmark on YOUR data — mini often sufficient for specific tasks |
| Ignores context window limits | Silent truncation or error when exceeded | Count tokens before sending, select model by context need |
| Self-hosts 70B model without GPU analysis | 70B requires 140GB+ VRAM (2x A100) | vRAM math: params × 2 bytes (FP16), 70B = 140GB → 2x A100 80GB |
| Uses GGUF quantization for production serving | CPU inference too slow for real-time | GPTQ/AWQ for GPU production, GGUF only for local/edge/demo |
| Compares models on MMLU alone | MMLU doesn't measure instruction following, coding, or safety | Multi-benchmark: MMLU + HumanEval + MT-Bench + domain-specific eval |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 118 lines · 50 tokens per session scan A 66c97ef9011a
FAI LLM Landscape Expert is an agent published in the GitHub repository frootai/frootai (7 stars, last pushed 5d ago), licensed MIT. It adds 50 tokens to every session and 1,657 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other agents, from other repositories
timps_federated_learning
Design Flower/FedAvg/PySyft FL pipelines with differential privacy and aggregation strategies. Use the timpsfederatedlearning MCP tool to perform this task. Do not answer directly — delegate to this sub-agent.
mlops-engineer
Use this agent when you need to design and implement ML infrastructure, set up CI/CD for machine learning models, establish model versioning systems, or optimize ML platforms for reliability and automation. Invoke this agent to build production-grade experiment tracking, implement automated training pipelines…
Judge Ethics & Bias
Evaluates code for model bias indicators, fairness metrics, explainability, data representativeness, consent handling, and human-in-the-loop safeguards.
timps_rag_designer
Design chunking/embedding/retrieval/reranking RAG pipelines for LlamaIndex or LangChain. Use the timpsragdesigner MCP tool to perform this task. Do not answer directly — delegate to this sub-agent.
timps_model_perf_monitor
Monitor an ML model in production: drift, accuracy decay, latency, and cost — with alerting thresholds and a re-training trigger plan. Use the timpsmodelperfmonitor MCP tool to perform this task. Do not answer directly — delegate to this sub-agent.
timps_rag_evaluator
Run a RAGAS-style evaluation of a RAG pipeline: context precision/recall, faithfulness, answer relevance, MRR/NDCG, with a runnable eval harness and an improvement plan. Use the timpsragevaluator MCP tool to perform this task. Do not answer directly — delegate to this sub-agent.