ModelsGemini 3.5 Flash: The Pricey Flash That Now Routes to 3.6
Gemini 3.5 Flash tripled the Flash price in May 2026 and was superseded two months later. What it delivered, why it cost so much, and what happens to calls that still name it.
ReadNotes on models, agents, and what inference costs
Practical writing for developers building with large language models — how they work, how to pick one, and how to keep the bill predictable.
Picking between models — and testing them on your own work.
Models
ModelsGemini 3.6 Flash: The Long-Context Workhorse and Its 2027 Price
Gemini 3.6 Flash writes fewer tokens than 3.5 Flash and roughly doubles its 1M-context recall. What it does well, where it trails, and what its price does in January.
Read
ModelsGemini 3.7 Flash: Three Weeks as the Newest Flash Model
Gemini 3.7 Flash shipped on 13 August 2026 and now redirects to 3.8 Flash. What it changed over 3.6, why the redirect matters, and how to stop model IDs moving under you.
Read
ModelsGemini 3.8 Flash: Frontier Agent Scores at a Flash Price, For Now
Gemini 3.8 Flash matches frontier models on some agent benchmarks at a fraction of the price, and spends tokens by design. Effort levels, migration notes and the 2027 price.
ReadModelsGemma 2 27B in 2026: When an 8K-Context Model Still Makes Sense
Gemma 2 27B was a strong open model in mid-2024. Today its 8K window and pricing make it a legacy pick. What it is, where it still fits, and how to migrate off it.
Read
ModelsGemma 3 12B: How Much of the 27B You Keep at Under Half the Size
Gemma 3 12B matches or beats the 27B on some of Google's own benchmarks and runs on an 8 GB laptop GPU at int4. Where the gap is real, and which jobs it suits.
ReadModelsGemma 3 27B: The Open Multimodal Model Sized for One GPU
Gemma 3 27B brought 128K context and vision to a model that fits a 24 GB card with QAT weights. The real memory maths, the benchmarks, and whether Gemma 4 replaces it.
ReadModelsGemma 3 4B: What a Small Multimodal Model Is Actually Good For
Gemma 3 4B reads images, holds 128K tokens and fits in under 3 GB at int4. It also scores 4.0 on SimpleQA. The jobs a 4B model does well, and the ones to keep from it.
ReadModelsGemma 4 26B A4B: A 3.8B-Active MoE Under Apache 2.0
Google's Gemma 4 MoE holds 25.2B parameters but computes with 3.8B per token, reads images and video, and drops the old Gemma licence. What that means in practice.
ReadModelsGemma 4 31B Guide: Apache 2.0, 256K Context, One Big GPU
Gemma 4 31B is a dense 30.7B model with image input, a 256K window and Apache 2.0 weights. What it does well, what it costs to run, and when the 26B MoE is better.
ReadModelsGLM-4.5 Air Guide: 12B Active, MIT Weights, One-Node Serving
GLM-4.5 Air is the 106B-total, 12B-active sibling of GLM-4.5. What it is, what hardware it needs, what it costs, and when it beats the full model.
ReadModelsGLM-4.5 Guide: The Release That Set Z.ai's Agentic Direction
GLM-4.5 is a 355B MoE with 32B active, hybrid thinking and MIT weights. What it introduced, what its benchmarks mean, and whether to run it over GLM-4.6.
ReadShowing 61–72 of 446 articles