Local AI on Apple Silicon
All Articles
74 articles
-
GPT-6.1 Terra, Pro and Batch: Four OpenRouter Choices Explained
GPT-6.1 Terra pricing, Pro reasoning, Batch processing, independently measured benchmarks, 272K-token pricing and EU routing.
-
Kimi K3.1 Pro: API Pricing, Benchmarks and Hermes on Mac
Kimi K3.1 Pro launched October 23: U.S. API pricing, million-token context, independent benchmark data, Hermes integration and open-weight status.
-
MiniMax M3.1 Flash: Pricing, Benchmarks and Mac Setup
MiniMax M3.1 Flash open weights released Oct 23, 2026: 428B/23B MoE, MLX for Mac, license restrictions, API pricing and coding benchmarks.
-
MiniMax M2.7 on Mac: API Pricing and the New M Plan
MiniMax M2.7 on Mac: API rates, M Plan vs legacy Token Plan, 5-hour/weekly windows, credits, safe billing, US privacy and Ollama Cloud.
-
DeepSeek V4.1 Flash 1017: Coding Gains, API Costs, and Agent Setup
DeepSeek V4.1 Flash 1017 launched on October 17, 2026: reported coding gains, official US-dollar API rates, cache economics, agent-integration pitfalls.
-
Claude Fable 5 Is Back: Legacy Status, API Pricing and Mac Alternatives
October 14, 2026 Claude Fable 5 legacy status, USD API pricing versus Fable 5.1, Opus 5.5 and Sonnet 5.5, 30-day retention and local Mac alternatives.
-
Meta Muse Wonder: Pricing, Benchmarks, API Access and Mac Setup
What Meta Muse Wonder really costs: the API, OpenRouter access, independent scores, 1M context, Wonder Full, privacy and Mac workflows.
-
Claude Fable 5.1 on Mac: Pricing, Benchmarks and API Changes
Claude Fable 5.1 in October 2026: $10/$50 API pricing, $0.25 cache reads, Opus 5.5 and Sonnet 5.5 comparison, breaking API changes, and OpenRouter on Mac.
-
Muse Spark 1.3 reviewed: benchmarks, pricing, API availability, and what changed from 1.2
Muse Spark 1.3: Meta Standard/Contributor pricing, cache rates, max reasoning, API limits and vendor benchmarks fact-checked.
-
Meta Muse Glimmer 30B fact-checked: 24/32 GB, DFlash, Qwen3.6-27B
Muse Glimmer 30B checked: 24/32 GB hardware, DFlash speed, 131K context, Ollama/MLX and independent benchmarks against Qwen3.6-27B.
-
Apple Intelligence Local AI: On-Device Models, PCC and Apple Silicon
Which Apple Intelligence requests stay on your Mac, which go to Private Cloud Compute or ChatGPT, and which need an M3 with 12GB.
-
Gemma 4 26B vs Qwen3.6 27B on Mac: RAM, Ollama and MLX
Gemma 4 26B A4B and Qwen3.6 27B compared on Mac: RAM, Ollama, MLX, coding, multimodality and the newer Qwen3.8 successor.
-
LM Studio vs Ollama: Which Is Better on Mac?
LM Studio or Ollama on Apple Silicon: GUI, CLI, local APIs, GGUF, MLX, RAG, offline operation and security compared.
-
Ollama on Mac mini M4: local AI setup, memory limits and the cloud trap
Set up Ollama on a Mac mini M4 and pick the right models per RAM tier: Gemma 4 and Qwen for 16–64GB, local API, Open WebUI, context length and the cloud trap.
-
Mac mini M4 for Local AI in 2026: 16GB, 24GB or M4 Pro?
The Mac mini M4 with 24GB is the practical pick for local AI: 4B-8B models with headroom. 32GB opens 8B-14B, M4 Pro adds bandwidth. Check Activation Lock used.
-
GLM-5.2 on Mac: Z.ai vs OpenRouter Pricing and Limits
GLM-5.2 has MIT-licensed weights and 1M context. Compare direct Z.ai rates, changing OpenRouter pricing, and practical Mac deployment limits.
-
Best MacBook for AI Development in 2026: M5 Pro 48GB vs 64GB
M5 Pro 48GB is the best overall MacBook for local AI development (7B-32B models). M5 Air 24GB, M5 Pro 64GB and M5 Max 128GB for Ollama, MLX, LM Studio.
-
Moondream2 on Mac: 1.7 GB Vision Without the Cloud
Run Moondream2 locally on Apple Silicon: Ollama setup, Python API, memory guidance, privacy and the limits of this compact vision model.
-
Gemma 3 on Mac: Variants, Vision and Memory Limits
Gemma 3 on Apple Silicon: which Ollama variants support text and images, what package size fits, and why 128K context is not a RAM promise.
-
Baidu ERNIE 5.1: strong cloud model, not a local Mac setup
Baidu ERNIE 5.1: listed in international Qianfan API with USD context-based pricing, account restrictions and practical Mac limitations.
-
How Much RAM Does a Local LLM Need on a Mac?
16GB runs 7B–8B models, 24–32GB is the sweet spot for 12B–32B, and 70B Q4 needs 64GB with checks. Mac RAM tiers for Ollama, LM Studio and MLX, incl. KV cache.
-
Best Open-Weight LLMs for Mac in 2026: RAM, Ollama Tags & Picks
Open-weight LLM picks for 8–64GB Macs: Gemma 4 for smaller systems, Qwen3.8-27B for 32GB+, Ollama tags, realistic memory margins and a fair speed test.
-
Apple Intelligence vs Local AI: Mac Privacy Guide
Apple Intelligence routes requests to Private Cloud Compute or ChatGPT, so it is not local inference. A local Ollama or MLX model needs cloud features off.
-
Google Gemini Agent at Work 2026: Features, Pricing and Availability
Google Gemini Agent at Work 2026: enterprise private preview, model routing with Claude, coworker agents, pricing limits and what Mac users can actually access.
-
Claude Sonnet 5.5: Pricing, Benchmarks, API IDs and Migration Changes
Claude Sonnet 5.5 analyzed: $2/$10 API pricing, 1M context window, current benchmarks, provider model IDs and breaking changes from Sonnet 5.
-
DeepSeek V4.1 Flash: API Price, 1M Context, Vision and Mac Reality
DeepSeek V4.1 Flash has 1M context, native vision, a 552B MoE backbone and very low API prices. Current V4 Pro status, Mac limits and open weights.
-
Grok 4.6 on Mac: API Pricing, 500K Context and the Desktop App
Grok 4.6 launch facts: confirmed API rates ($2/M input, $6/M output), 500K context, the macOS app, and what xAI has not disclosed yet.
-
Gemini 3.7 Flash on Mac: API Price and Migration to 3.8
Gemini 3.7 Flash deprecated October 8: 3.8 routing, current rates, working API examples and historical benchmarks on Mac.
-
Qwen3.8-Max API: $1.65 US Global vs $2 Singapore pricing
Qwen3.8-Max: US Virginia Global API at $1.65/$4.951 vs Singapore International $2/$6 per million tokens, weights and Mac limits.
-
Grok on Mac: Grok Bot Desktop App, Grok Build and API
Grok for Mac: Grok Bot, Grok Build, and xAI API. Grok 4.5 vs 4.7 pricing, caching, U.S. regional endpoint, and cloud-only limitations.
-
Nex N2 Pro on Mac: What 397B MoE Means in Practice
Nex N2 Pro on Mac: 397B/17B MoE, official 400GB FP8 release, realistic memory limits, and secure remote use without invented MLX benchmarks.
-
Gemma 4 12B on Mac: RAM, Ollama, MLX and Multimodal Support
Gemma 4 12B runs on 16GB Macs in 4-bit (~8GB package): 256K context, image, audio and video input, plus a repeatable 16/24/32GB RAM check for Ollama and MLX.
-
NVIDIA Nemotron 3 Ultra on Mac: Cloud Tag, 256K and NIM
Use Nemotron 3 Ultra from a Mac without confusing cloud access with local inference: Ollama Cloud, native 256K, optional NIM 1M, real hardware limits.
-
Gemini 3.5 Flash on Mac: How to Use It, Pricing, and Local Alternatives
What Gemini 3.5 Flash can do, how to use it from a Mac, what it costs, how Google handles your data, and which local models fit when you need offline AI.
-
Can Gemini 3.5 Flash Run Locally on Mac? Ollama, MLX & Pricing
Gemini 3.5 Flash deprecated: switch to 3.6 Flash, current API token rates and model IDs, privacy and offline Mac alternatives.
-
Claude Opus 4.8 on Mac: Legacy API vs Opus 5.5
Opus 4.8 vs. Opus 5.5 on Mac: verified US API prices, cache accounting, breaking agent migration changes, and executable test protocol.
-
Mistral Large 4: 1.05T Parameters – Can Le Chonk Fit on a Mac?
Mistral Large 4 preview checked against conflicting official 49B vs 52B active-parameter counts, 1.05T storage, 1M context and temporary API sale pricing.
-
Reflection AI Beam: 501B Open-Weight Model, Benchmarks, Availability, and Mac Feasibility
Reflection AI Beam: 501B total parameters, 23B active, benchmarks, 1M effective context, release status, API caveats and theoretical Mac feasibility.
-
DeepSeek Harness Explained: What It Is, What You Get, and What You Need
DeepSeek Harness is an open agent runtime with interchangeable models, tools, MCP, subagents and scheduling. Here is what it includes, costs and requires.
-
Pareto 26.9 vs 26.10 Preview: Prices, API and Mac Use
Pareto 26.9 combines several LLMs behind one API. A source-checked guide to pricing, benchmarks, privacy, OpenRouter and Mac usage.
-
Qwen3.8-Flash Explained: Pricing, 1M Context, Benchmarks, and Flash-Next
A source-checked guide to Qwen3.8-Flash pricing, its 1M context window, API features, benchmarks, and the separate Flash-Next open-weight release.
-
GLM-5.3: API pricing, open weights and what a Mac needs
For almost every Mac, GLM-5.3 is an API model: $1.40/$4.40 per 1M tokens at Z.ai. The 753B open weights need 217-467GB, so local runs need a Mac Studio.
-
NVIDIA Nemotron 3.5 Lightning: 30B agent model, 1M context, RTX & Mac
NVIDIA Nemotron 3.5 Lightning: 30B MoE with 3B active, up to 1M context, $0.05/$0.20 per 1M tokens on OpenRouter, about 670 tok/s on NVFP4.
-
MiniMax H3 on Mac: US Weights License, API Prices & H3-Max
MiniMax H3 generates AI video up to 2K. API costs: $0.08/s at 768P, $0.13/s at 2K. The open-weights license excludes US self-hosting; hosted API works.
-
Poolside Laguna XS.2 on Mac: Open-Weight Coding Model, Benchmarks and RAM
Poolside Laguna XS.2 vs XS 2.1 on Mac: Ollama Q4 and MLX tags, context, Metal chat failures, licenses and safe agent migration.
-
StepFun Step 3.7 Flash on Mac: 198B MoE, 256K Context and the Local Reality
StepFun Step 3.7 Flash for U.S. Mac users: 198B MoE weights, 128GB RAM recommendation, direct vs OpenRouter API rates, caching, and privacy.
-
Qwen3.7-Max OpenRouter Pricing: 1M Context, API Setup & Mac Limits
Qwen3.7-Max on Mac: OpenRouter versus Alibaba US-scope prices, cache billing, safe API calls and local alternatives.
-
Qwen3.6 on Mac: 27B, 35B-A3B, Vision and Ollama
Run Qwen3.6 27B or 35B-A3B on Mac: exact Ollama tags, memory headroom, local vision with Python, context settings and measured API checks.
-
Claude Haiku 5.5: pricing, benchmarks and 1M context
Claude Haiku 5.5 launched on October 7, 2026. Verified API pricing, 1M context, migration changes, benchmark context and a comparison with Haiku 4.5.
-
Google Gemini Pricing 2026: U.S. Plus, Pro, Ultra and API Costs
Current U.S. Gemini pricing for AI Plus, AI Pro and AI Ultra, the 2026 student offer, October model-access changes and Gemini API token rates.
-
OpenAI DevDay 2026: Dots, GPT-6.1 Sol, Pro 500 & 25 Launches
OpenAI DevDay 2026: GPT-6.1 Sol, Dots rollout including EEA restrictions, Pro 500 and what changed with GPT-6 Chat on October 7.
-
TypeSafe Jev Router: How `typesafe/jev-router` Chooses Models, Reasoning and Cache Trade-offs
A source-checked guide to Jev Router: 1M context, model and reasoning routing, cache-aware switching, OpenRouter API setup, ZDR, benchmarks and limitations.
-
Ternary Bonsai 2 27B fact check: 5.9 GB, 262K context, and the runtime catch
Ternary Bonsai 2 27B: 5.95 GB GGUF needs PrismML's fork, MLX requires a bundled loader, and Mac RAM and vision limits need checking.
-
GPT-6 Astra 2026: API Price, 1.05M Context and ChatGPT Access
GPT-6 Astra costs $10/$50 per 1M tokens, has a 1.05M context window and 128K output. Current API, ChatGPT, Codex and Mac details.
-
Gemini 3.8 Flash: pricing, API limits, benchmarks and 3.7 migration
Gemini 3.8 Flash costs $0.75 per 1M input and $3.75 per 1M output tokens through Dec 31, 2026, then $1.50/$7.50. 1M context, limits, benchmarks.
-
Tencent Hy4 Preview: US API Routing, 770B, Context & Costs
Tencent Hy4 Preview for U.S. developers: 131K vs 64K output-token provider limits, USD routing prices, Chinese Tencent Cloud billing, Apache 2.0 and Mac memory.
-
Qwen3.8-27B on Mac: MLX setup, memory and benchmarks
Qwen3.8-27B landed August 14, 2026 as an official open-weight checkpoint. Metadata, MLX/GGUF ecosystem, Mac memory math and remaining benchmark gaps.
-
Meta Muse Spark 1.1 on Mac: What Changed by Muse Spark 1.3
Muse Spark 1.1 remains listed on Meta Model API. Compare 1.3, provider access, Standard vs Contributor pricing and Mac data flows.
-
Tencent Hy3 on Mac: OpenRouter, 295B MoE, Apache 2.0 and Local Limits
Tencent Hy3: 295B MoE, 21B active, Apache 2.0. Why it stays a cloud model on Mac, what OpenRouter (incl. the free endpoint) costs, and its privacy limits.
-
Sakana Fugu Ultra: An AI Orchestrator, Not a Model You Can Download
Sakana Fugu Ultra v2 vs Max for U.S. Mac users: real API bills with orchestration, cache and 272K surcharge, tested pricing math and safe setup.
-
Gemma 4 on Mac: Variants, Modalities and Memory Planning
Gemma 4 on Mac: verified Q4 tags, memory planning and a reproducible Ollama API benchmark protocol with real token-per-second metrics.
-
Gemini 4 Argon: 1M Output, $2/$10 Pricing & Benchmarks
Google confirms 1M output tokens, $2/$10 introductory pricing and new DeepMind benchmarks for Gemini 4 Argon. The 2M context leak remains unconfirmed.
-
Claude Opus 5.5: Pricing, Benchmarks, API, and What Actually Changed
Claude Opus 5.5 pricing, benchmarks, API model ID, safety changes and original cost calculations versus Opus 5.
-
GPT-6 Sol vs Luna: Pricing, Pro Mode, Batch and Every Difference
GPT-6 Sol vs Luna: API pricing, 1.05M context, Pro reasoning, Batch discounts, cache costs and the eight visible variants explained.
-
Grok 4.7: pricing, benchmarks, 500k context and API details checked
Grok 4.7 fact check: release, 500k context, API pricing from $2/M input, long-context surcharge above 200k tokens, benchmarks and independent measurements.
-
Xiaomi MiMo-V2.6: Pro, Flash and UltraSpeed fact check
Xiaomi MiMo-V2.6 Pro, Flash and UltraSpeed: correct API cost calculations, cache rates, 1M context and Responses API limits.
-
DeepSeek V4 Pro 0813: Pricing, Benchmarks and the Mac Reality Check
DeepSeek V4 Pro 0813 versus V4.1 Flash in October: U.S. API rates, cache billing, model aliases, original agent benchmarks, and Mac limits.
-
Qwen3.6-27B Fable Fusion 711 on Mac: GGUF Sizes and RAM Choices
Fable Fusion 711 on Mac: start with the 18.0 GB Q4_K_M GGUF on 32 GB. File sizes for 24 to 64 GB, MTP and vision files, no original Mac benchmarks.
-
Meituan LongCat-2.0: The 1.6-Trillion-Parameter Model Explained
LongCat-2.0 fact check: 1.6T MoE weights, 1M context, verified USD API pricing, cache scenarios and LongCat-2.5-Preview comparison for Mac users.
-
Kimi K3 on Mac: Open weights are here — why local use is still impractical
Kimi K3 on Mac: 2.8T parameters, 1M context, 16/896 experts. Current U.S. API pricing, cache writes, access and license explained.
-
Claude Sonnet 5 on Mac: Agents, Coding, 1M Context and API Costs Explained
Sonnet 5 after the Sonnet 5.5 release: $2/$10 USD API list rates, 1M context, native model IDs, macOS workflows and migration advice.
-
Gemini 3.1 Flash Lite Image on Mac: Nano Banana 2 Lite Explained
Gemini 3.1 Flash Lite Image October guide: Nano Banana 2.1 comparison, US 1K API pricing, Batch, working generateContent REST example and Mac limits.
-
macOS 27 Golden Gate compatibility: Does it run on your Mac? Intel support ends
macOS 27 Golden Gate supports no Intel Mac: only Apple silicon from M1, plus the A18 Pro MacBook Neo. Supported-Mac list, 1-minute check, M3 + 12GB rule.
-
Kimi K2.7 Code on Mac: Cloud, API, or GGUF?
Kimi K2.7 Code on Mac: U.S. API rates versus K3, cloud/Ollama separation, Kimi Code subscriptions, 256K versus 1M context, and GGUF hardware.