Model Selection

MoE models overfit repeated training data earlier than dense models, with total parameter count driving the effect. Strong masking helps, but unique data remains the stronger baseline.

For edge vision, specialized training outweighed model size: 300M-parameter BioCLIP beat 2–8B VLMs, while field imagery degraded every model and open-set prompts produced invented species.

GPT-6 Astra combines reasoning, computer use, and writing and design judgment, potentially widening the range of knowledge-work tasks one agent can handle.

DeepSeek V4.1 Flash brings vision, tool use, reasoning, and prompt caching to Vercel AI Gateway, with direct setup paths for Claude Code, Codex, and Cursor.

Vercel AI Gateway adds GPT Image 2.5 Flare for faster iteration and Sunburst for tighter composition control. Both support generation, reference-based edits, and transparent backgrounds.

Molecular benchmark accuracy can reflect retrieval of published values rather than prediction. Contamination audits should test digit-level recall and repeat runs across reasoning settings.

A compact classifier distilled from LLM rationales handled product-pair decisions without inference-time LLM calls. Category adapters improved accuracy further while retaining large speed and cost gains.

Vercel AI Gateway now serves the medical-focused Ling 3.0 Flash Sante free through October 4. Use its free-only model ID to prevent requests from converting to paid usage afterward.

On-policy distillation recovered most full-data gains from one query; diverse rollouts mattered more than dataset size, while slow student alignment remained the bottleneck.

Legora says GPT-6 Astra reviewed 41 documents within minutes, caught every planted error, and improved its workflow result by nearly 40%, though the underlying measure is unspecified.

Playco reports that GPT-6 Astra halved manual fixes while producing three themed game prototypes from one grey-box base, suggesting less cleanup in model-driven iteration.

OpenAI introduces GPT-6 Astra with claimed advances in computer use, coding, cybersecurity, and science. The supplied material gives no benchmarks or implementation details to assess those gains.

GPT-6 Astra is OpenAI's first model rated Critical for cybersecurity capability under its Preparedness Framework, a material consideration for security-sensitive agent access and controls.

A coding-specialized model paired post-training with an iterative generate-evaluate-refine loop to exceed the top IOI 2026 human score. The reusable idea is feedback-driven test-time search.

A UE5M3 block-scaling recipe trained an 8B model in FP4 without Hadamard transforms or BF16 final layers, while reporting better losses and downstream estimates than the compared recipe.

Muse Spark 1.3 gives coding agents a 1M-token, multimodal model through Vercel, with a cheaper contributor tier that permits Meta to train on submitted inputs and outputs.

Gemini 3.8 Flash brings multimodal input, tool calling, web search, and default reasoning to coding agents through Vercel. Its temporary 50% discount runs through December 31.

When quantizing an open model, spend a small extra precision budget across the network before protecting a few “important” layers; causal tests found the damage was usually diffuse.

Small proxy models may be enough to choose an SFT-versus-RL annotation split: the paper finds broad near-optimal ranges that transfer to larger models.

A production-derived post-training recipe consolidated more than 200 internal apps onto one self-hosted model by training separate experts for distinct quality gaps, then merging them.

Claude Fable 5.1 reaches Vercel AI Gateway with ordered fallbacks for classifier refusals, but its 30-day retention policy rules out zero-data-retention workloads.

Vercel’s gateway now exposes a pinned Qwen snapshot aimed at larger coding projects and longer agent runs, with a dated model ID that prevents silent upgrades.

A controlled ontology-learning study finds model size is a weak selector on its own. Dense 27B models beat larger sparse models on one task, while MoE models led another.

Adobe’s prototype assembles intent-specific page blocks from existing site content in roughly a second, making model latency and per-site evaluation part of frontend architecture.

Agent architectures are expiring quickly. Keep model, search, and orchestration choices replaceable, and evaluate platforms by how well they handle repeated change.

Tencent’s Hy4 Preview is now callable through Vercel AI Gateway and selectable in coding agents, adding an open MoE option with a 1M-token context window.

Three ways to combine RLVR domain experts perform similarly on average but diverge by task. Choose Merge for cheap reuse, Mix RL for training from pooled data, and MOPD for preserving expert gains.

Ling 3.0 Flash Fin adds a finance-focused reasoning and tool-calling option to AI Gateway, with separate model IDs for automatic billing or a hard stop after the free period.

Qwen 3.8 Flash is now selectable in Vercel AI Gateway and coding agents, with text-and-image input, a 1M-token context window, and responses up to 65k tokens.

GLM 5.3 Flash joins Vercel AI Gateway with text and vision input, a 1M-token context window, function calling, structured output, and streaming.

Wan 3.0 gives AI Gateway one video model ID for text, image, frame, and reference workflows, with async renders up to 30 seconds at 1080p and synchronized audio.

AI Gateway offers temporary free routes for MiniMax M3 and M2.7, but the -free model IDs hard-fail after September 6; standard IDs preserve provider fallback at normal rates.

Per-task model routing cut the demonstrated coding session’s cost from 44¢ to 14¢ with similar completion time, but builders still need workload-specific evals to validate quality.

Self-refinement pipelines need not use equally capable models: invest capacity in generation and revision, while a small critic may preserve gains at lower compute cost.

CLEAR conditionally activates a safety adapter instead of applying safety tuning to every prompt, reducing harmful completions while limiting benign-task degradation.

Lease End replaced a fine-tuned intent classifier with skills and runtime context, cutting production fixes from about a week to under an hour. Higher API spend was offset by lower maintenance cost.

TokEval links tokenizer properties to language, math, and code performance, offering cheaper screening signals before committing compute to pretraining sweeps.

A single-GPU SBERT beat the reported zero-shot LLM and vendor baseline for invoice coding, suggesting narrow, private classifiers can outperform broader models with modest local data.

GLM 5.3 is available through Vercel AI Gateway for coding agents, retaining a 1M-token context window while claiming better long-horizon engineering with fewer output tokens.

Vercel cut GPT-5.6 Sol pricing by 50% through September 18, making direct AI Gateway runs cheaper across every tier without changing model IDs or agent configs.