--- author: "Umesh Malik" canonical: "https://umesh-malik.com/blog/category/llm-engineering" description: "Explore LLM Engineering articles and tutorials by Umesh Malik — AI Engineer, LLM & GenAI Developer. Learn LLM Engineering best practices, practical tips, and in-depth guides." title: "Umesh Malik's Blog - LLM Engineering Articles | LLM Engineering Tutorials" tokens: 2670 generator: "scripts/generate-page-markdown.mjs" --- [← Back to Blog](https://umesh-malik.com/blog) # LLM Engineering 20 articles [![Qwen3.8 27B VRAM budget: FP8 weights plus KV cache at 262K context on a single GPU](https://umesh-malik.com/blog/qwen3-8-27b-vram-kv-cache-math-cover.png) LLM Engineering • Aug 15, 2026 ### Qwen3.8 27B VRAM: how to fit 262K context in 16 GiB, not 64 Qwen3.8 27B VRAM math: 25.9 GiB of FP8 weights plus 16 GiB of KV cache at 262,144 tokens, not 64. The arithmetic, and where a 48 GB card breaks. 9 min read Read more →](https://umesh-malik.com/blog/qwen3-8-27b-vram-kv-cache-math) [![How a paravirtualized Metal device reporting Apple GPU family 5 makes llama.cpp disable its simdgroup matrix kernels, and the throughput that returns when the guest reports family 9](https://umesh-malik.com/blog/fix-slow-llm-inference-macos-vms-cover.png) LLM Engineering • Aug 11, 2026 ### Fix slow LLM inference in macOS VMs: 12.6 → 207 tok/s LLM inference in macOS VMs collapses to 12.63 tok/s because the guest reports GPU family 5 and llama.cpp drops its matrix kernels. The check, and its limits. 9 min read Read more →](https://umesh-malik.com/blog/fix-slow-llm-inference-macos-vms) [![Timeline comparing the knowledge-cutoff date vendors claim for Claude Opus 5 against the earlier cutoff its answers actually reveal](https://umesh-malik.com/blog/testing-llm-knowledge-cutoffs-opus-5-cover.png) LLM Engineering • Aug 11, 2026 ### How to Test an LLM's Knowledge Cutoff: Opus 5's May Claim Falls Short Here's how to test an LLM's knowledge cutoff with three reproducible probes — the method showing Opus 5 claims May 2026 but answers like January 2026. 7 min read Read more →](https://umesh-malik.com/blog/testing-llm-knowledge-cutoffs-opus-5) [![Cover showing the vLLM VRAM budget split into model weights, runtime overhead and KV cache, with the KV cache block math that converts free VRAM into concurrent sequences](https://umesh-malik.com/blog/vllm-throughput-tuning-flags-cover.png) LLM Engineering • Aug 8, 2026 ### vLLM throughput tuning: configure these four flags, not a bigger GPU vLLM throughput tuning starts with KV cache blocks, not a bigger GPU. The four flags that decide your tokens/sec, and the one that quietly backfires. 10 min read Read more →](https://umesh-malik.com/blog/vllm-throughput-tuning-flags) [![Reinforcement fine-tuning: a 4B open model matching a frontier LLM on retrieval at a fraction of the cost](https://umesh-malik.com/blog/reinforcement-fine-tuning-small-models-retrieval-cover.png) LLM Engineering • Aug 6, 2026 ### Reinforcement Fine-Tuning: When a 4B Model Beats GPT-5.6 Reinforcement fine-tuning let a 4B open model match GPT-5.6 Sol on retrieval at 100x lower cost. How RFT works, and when it beats prompting a frontier LLM. 10 min read Read more →](https://umesh-malik.com/blog/reinforcement-fine-tuning-small-models-retrieval) [![Diagram of an LLM eval framework pipeline: tasks and configs produce runs, graders apply checks to produce grades](https://umesh-malik.com/blog/llm-eval-framework-smevals-cover.png) LLM Engineering • Aug 4, 2026 ### LLM Eval Framework: Grade Prompts, Models and Harnesses An LLM eval framework turns vibes into scores. How smevals structures tasks, configs, runners and graders — and how to ship your first eval today. 10 min read Read more →](https://umesh-malik.com/blog/llm-eval-framework-smevals) [![Layer-by-layer LLM inference streaming one transformer layer at a time from disk onto a 4GB GPU](https://umesh-malik.com/blog/run-70b-llm-on-4gb-gpu-airllm-cover.png) LLM Engineering • Aug 4, 2026 ### Run 70B LLM on 4GB GPU: AirLLM's Real Tradeoff Run 70B LLM on 4GB GPU hardware with AirLLM's layer-by-layer inference. The VRAM math is real — you just pay for it in disk bandwidth. The honest tradeoff. 11 min read Read more →](https://umesh-malik.com/blog/run-70b-llm-on-4gb-gpu-airllm) [![DeepSeek V4 Flash 0731 agent benchmark results compared against the preview build and V4-Pro Preview](https://umesh-malik.com/blog/deepseek-v4-flash-0731-benchmarks-cover.png) LLM Engineering • Aug 1, 2026 ### DeepSeek V4 Flash 0731 Benchmarks: 13B Active Beats 1.6T DeepSeek V4 Flash 0731 benchmarks: same 284B/13B architecture as the preview, re-post-trained only — and it beats the 1.6T V4-Pro Preview on nine agent tests. 9 min read Read more →](https://umesh-malik.com/blog/deepseek-v4-flash-0731-benchmarks) [![Cover for the Claude Opus 5 guide: benchmarks, the $5/$25 cost math, effort levels, and the two API breaking changes in the Opus 4.8 migration](https://umesh-malik.com/blog/claude-opus-5-guide-cover.png) LLM Engineering • Jul 24, 2026 ### Claude Opus 5 Migration: The Two API Changes That Break Your Code Two API changes break a Claude Opus 5 migration, and the effort parameter needs re-sweeping, not reusing. The fixes, the real cost math, and the benchmarks. 12 min read Read more →](https://umesh-malik.com/blog/claude-opus-5-guide) [![Kimi K3 vs Claude Fable 5 head-to-head showing benchmarks, pricing, and where the open 2.8T model wins](https://umesh-malik.com/blog/kimi-k3-vs-fable-5-cover.png) LLM Engineering • Jul 19, 2026 ### Kimi K3 vs Claude Fable 5: When the Open Model Is Worth the Switch Kimi K3 beats Claude Fable 5 on cost by a wide margin and loses on agentic tasks. The benchmarks that decide it, and how to run K3 where it actually wins. 6 min read Read more →](https://umesh-malik.com/blog/kimi-k3-vs-claude-fable-5) [![GPT-5.6 Sol vs Terra vs Luna comparison showing price, coding strength, and cost per task](https://umesh-malik.com/blog/gpt-5-6-sol-terra-luna-cover.png) LLM Engineering • Jul 11, 2026 ### GPT-5.6 Sol vs Terra vs Luna: The Routing Strategy That Cuts Cost GPT-5.6 Sol vs Terra vs Luna compared on price, coding, latency, and cost per task — plus a routing strategy that cuts your bill without wrecking quality. 5 min read Read more →](https://umesh-malik.com/blog/gpt-5-6-sol-vs-terra-vs-luna) [![OpenAI GPT-5.6 family showing Sol, Terra, and Luna tiers with benchmarks, pricing, and 1.05M context](https://umesh-malik.com/blog/gpt-5-6-cover.png) LLM Engineering • Jul 11, 2026 ### GPT-5.6 API: Pricing, Thinking Modes, and the Shared Context Trap GPT-5.6 API pricing ($1-$30/1M), the Ultra and Max thinking modes, and a 1.05M context window that is shared — with the fine print that breaks agent loops. 10 min read Read more →](https://umesh-malik.com/blog/openai-gpt-5-6-sol-terra-luna-guide) [![Cover for the Claude Fable 5 deep-dive guide: capabilities, the $10/$50 cost math, API behavior, and when to use Anthropic's most capable model](https://umesh-malik.com/blog/claude-fable-5-guide-cover.png) LLM Engineering • Jul 2, 2026 ### Claude Fable 5: Capabilities, Cost & When to Use It (2026) Claude Fable 5 is Anthropic's most capable model yet. The honest deep-dive: capabilities, the $10/$50 cost math, API behavior, and when to use it. 12 min read Read more →](https://umesh-malik.com/blog/claude-fable-5-guide) [![Cover for the Claude Sonnet 5 guide: benchmarks, pros and cons, use cases, and cost math versus Sonnet 4.6 and Opus 4.8](https://umesh-malik.com/blog/claude-sonnet-5-guide-cover.png) LLM Engineering • Jul 1, 2026 ### Claude Sonnet 5 for Coding: The Tokenizer Change That Moves Your Bill Claude Sonnet 5 for coding lands near Opus level at Sonnet prices — until the tokenizer change moves your bill. The real cost math, and when it beats Opus 4.8. 8 min read Read more →](https://umesh-malik.com/blog/claude-sonnet-5-guide) [![ChatGPT adult mode cover showing age prediction, adult verification, text-only scope, and safety guardrails](https://umesh-malik.com/blog/chatgpt-adult-mode-cover.png) LLM Engineering • Mar 16, 2026 ### ChatGPT Adult Mode: Is It Live Yet? (Status Explained) ChatGPT adult mode is still delayed — OpenAI's official status, what the feature would allow, why it was pushed back, and answers for parents. 7 min read Read more →](https://umesh-malik.com/blog/chatgpt-adult-mode-delay-guide) [![ChatGPT interactive learning cover showing math and science concepts becoming visual and interactive inside ChatGPT](https://umesh-malik.com/blog/chatgpt-learning-visuals-cover.png) LLM Engineering • Mar 12, 2026 ### ChatGPT Interactive Math and Science Visuals: What to Know ChatGPT interactive math and science visuals launched in March 2026: how the new learning modules work, who gets access, and why students benefit. 6 min read Read more →](https://umesh-malik.com/blog/chatgpt-interactive-math-science-visuals-guide) [![OpenAI GPT-5.4 overview showing professional work, coding, computer use, and 1M context](https://umesh-malik.com/blog/gpt-5-4-cover.png) LLM Engineering • Mar 6, 2026 ### GPT-5.4 for Agents: Computer Use, MCP Tool Calls, and Real Pricing GPT-5.4's native computer use and MCP tool calls are the real upgrade for agents. What holds up in a loop, what the 1M context costs, and how Pro compares. 12 min read Read more →](https://umesh-malik.com/blog/openai-gpt-5-4-complete-guide) [![OpenAI GPT-5.3 Instant overview showing three key improvements: fewer refusals, better web answers, and smoother conversational tone](https://umesh-malik.com/blog/gpt-5-3-instant-cover.png) LLM Engineering • Mar 4, 2026 ### OpenAI GPT-5.3 Instant: 26.8% Fewer Hallucinations, Reduced Refusals, and Better Web Answers GPT-5.3 Instant brings 26.8% fewer hallucinations, fewer needless refusals, and better web-sourced answers — what changed and why it matters for devs. 10 min read Read more →](https://umesh-malik.com/blog/openai-gpt-5-3-instant-fewer-refusals-better-answers) [![DeepSeek V4 launch preview showing AI model race between China-first chips and U.S. rivals](https://umesh-malik.com/blog/deepseek-v4-release-cover.png) LLM Engineering • Mar 1, 2026 ### DeepSeek V4 vs US AI Models: Benchmarks, Architecture, and What It Means for the Industry DeepSeek V4 is expected in early March 2026. Here is what is confirmed, what remains unverified, and how it challenges U.S. AI rivals. 10 min read Read more →](https://umesh-malik.com/blog/deepseek-v4-release-challenge-us-ai-rivals) [![The Seedance 2.0 crisis: Hollywood confrontation with AI-generated video showing film reel colliding with AI](https://umesh-malik.com/blog/seedance-cover.png) LLM Engineering • Feb 16, 2026 ### Seedance 2.0: The Two-Line Prompt That Broke Hollywood ByteDance's Seedance 2.0 made a photorealistic Tom Cruise vs Brad Pitt fight from a two-line prompt — igniting Disney, Paramount, and SAG-AFTRA backlash. 13 min read Read more →](https://umesh-malik.com/blog/seedance-2-hollywood-ai-copyright-crisis)