ModelsReMM SLERP L2 13B: Rebuilding MythoMax From Its Parts
ReMM SLERP is a 2023 Llama 2 13B merge that rebuilds MythoMax from newer ingredients. The recipe, the non-commercial licence, and when it is worth calling.
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
ModelsSakana Namazu: Kimi K2.6 Retrained for Japanese Business Work
Sakana Namazu takes Kimi K2.6 and post-trains it for Japanese language, business norms and neutrality, with built-in search and code execution. What it changes.
Read
ModelsSchematron V2 Small Guide: HTML to JSON With a Specialist
Schematron V2 Small is a small model from Inference.net that does one job: turn HTML into JSON that matches your schema. How to call it, and when it beats a big model.
Read
ModelsSchematron V2 Turbo: Cheap HTML-to-JSON Extraction at Volume
A small extraction model that takes a JSON schema instead of a prompt. What Turbo trades against V2 Small, what a page costs, and how to prepare HTML for it.
Read
ModelsSeed 1.6 Flash Guide: Cheap Multimodal Input, Short Shelf Life
Seed 1.6 Flash is ByteDance's low-cost 256K multimodal model with optional thinking. Where its price shape pays off, and what to move to before it retires.
ReadModelsSeed 1.6 Guide: Adaptive Thinking and a Retirement Date
ByteDance Seed 1.6 decides per prompt whether to think. How that works, what the 230B MoE costs to call, and why its OpenRouter endpoint needs a migration plan.
Read
ModelsSeed 2.0 Code Guide: A Coding Variant With No Scorecard
ByteDance tuned Seed 2.0 Code for programming and agent loops but published no coding benchmarks for it. What is known, what it costs, and how to evaluate it.
ReadModelsSeed 2.0 Lite Guide: Two Builds, One Name, a Sensible Default
Seed 2.0 Lite shipped in February and was rebuilt in April with audio and GUI gains. What changed between builds, how close it gets to Pro, and when to choose it.
ReadModelsSeed 2.0 Mini Guide: Throughput First, Reasoning on a Dial
Seed 2.0 Mini is ByteDance's high-concurrency tier: 256K context, video input and four effort levels for $0.10 in. Where it holds up and where it falls away.
ReadModelsSeed 2.1 Turbo Guide: Where It Keeps Up With Pro and Where It Slips
ByteDance's cheaper Seed 2.1 model takes text, images and video. What the official Turbo-versus-Pro table shows, what it costs, and the jobs where the gap bites.
ReadModelsSkyfall 36B V2 Guide: An Upscaled Mistral Small for Fiction
Skyfall 36B V2 grows Mistral Small 2501 to 36B and trains it for storytelling. What an upscale buys, why 32K context limits it, and when to choose it over Cydonia.
ReadModelsSolar Mini 4: Upstage's 3B-Active Model With a 512K Window
Solar Mini 4 is a 35B mixture-of-experts with 3B active, a 512K context and a reasoning mode. What it costs, where it fits, and when Solar Pro 4 is worth more.
ReadShowing 301–312 of 446 articles