TokenBlog

Model releases, pricing breakdowns, and practical guides for developers.

Model ReleaseJune 2, 2026·8 min read

MiniMax M3 claims GPT-5.5-class coding for a tenth of the price. The benchmarks are self-reported and the cheap tier stops at 512K.

MiniMax M3 went live on June 1 at $0.60 input and $2.40 output per million tokens, with a 50% launch promo cutting it to $0.30/$1.20 for the first week. It pairs a 1M context window and image-plus-video input with coding scores MiniMax says match GPT-5.5 on SWE-Bench Pro. Two catches: every benchmark is vendor-reported, and the cheap rate doubles once a request crosses 512K input. We work through the tiered rate card, the cost math against GPT-5.5, Opus 4.8, and DeepSeek V4-Pro, and where M3 actually wins.

TokenBlog

MiniMax M3 claims GPT-5.5-class coding for a tenth of the price. The benchmarks are self-reported and the cheap tier stops at 512K.

GPT-5.5 has a pricing cliff at 272K tokens. Cross it and the whole request bills double.

Cohere shipped a 218B Apache 2.0 model that runs on two H100s. The hosted rate matches Command A; the license is the actual news.

Anthropic kept the Opus list price flat for the third release running. The headline is the new Fast mode: $10/$50 at 2.5x speed.

Qwen3.7 Max ties Opus 4.7 on intelligence and beats it on agentic coding, at half the input price. The catch: closed weights and no images.

Cursor Composer 2.5 lands one point behind Opus 4.7 on SWE-Bench and bills ten times less. The promo just ended.

GPT-5.5 Pro lists at $30/$180 per million tokens, and it drops the cache discount that keeps the standard model cheap.

Grok Build runs coding agents at $1 in, $2 out. The catch is that xAI published zero benchmarks.

DeepSeek scrapped the May 31 price cliff. The 75% V4-Pro cut is the permanent rate now.

Google cut AI Ultra to $100 at I/O. Now all three AI giants charge the exact same $20, $100, $200 ladder.

Your Claude subscription stops paying for agents on June 15. The SDK and claude -p move to a separate credit billed at full API rates.

Gemini 3.5 Flash costs 3x what Gemini 3 Flash did. Google priced it that way because it beats their own Pro model on agentic work.

xAI quietly swapped Grok 4.3's weights for Colossus-2-trained ones and switched on Agent Mode. Same endpoint, same $1.25 input, new build.

SubQ 1M-Preview shipped 12M tokens of context, no public rate card, and a $8-vs-$2,600 cost story that does not reconcile.

Qwen3.6 Max Preview is the first Qwen flagship Alibaba shipped closed. It took SWE-Bench Pro, lost SWE-Bench Verified by omission, and costs 4x what Qwen3.6 Plus does.

Anthropic is raising at $950B, up 2.5x in three months. Every recent move on the Claude price sheet has been a cut. That gap is the whole story.

MiniMax M2.7 lists at $0.30 input, $1.20 output. The GDPval-AA score is the highest open-weight number on the board. The SWE-Bench Verified column is empty.

Qwen3 Coder Next is still $0.11 per million input. Kimi K2.6 costs 7x more, GPT-5.5 costs 41x more, and the benchmark gaps do not justify either.

Mercury 2 outputs at 788 tokens per second for $0.75 per million. The diffusion math turns frontier reasoning pricing into a rounding error.

OpenAI's three new voice models price three different ways. Here is what an hour actually costs.

Tencent's Hunyuan HY3 Preview is the cheapest frontier-class model, and it's 14 points behind the leaders on coding

GPT-5.5 vs Opus 4.7 vs Gemini 3.1 Pro: the cheapest one depends entirely on whether you cross 200K tokens

DeepSeek V4-Pro's 75% promo ends May 31. After that, the price is 4x what most people are quoting.

GLM-4.7-flash sits at $0.07 input on AWS Bedrock and Vertex. Most coverage skipped this one.

Mistral Medium 3.5 charges 17x more than DeepSeek V4 Flash and loses the only benchmark they both report

Three weeks of Opus 4.7 bills are in. The tokenizer change costs an extra 25 to 37 percent in production.

GPT-5.5 doubled the per-token price. The 40% efficiency claim does not get OpenAI back to even.

GLM-5.1 took SWE-Bench Pro at $1.40/M input. The catch is the small print, not the price.

Grok 4.3 quietly cut prices 40-60%, then shipped a voice cloner. The pricing math is now hard to ignore.

DeepSeek V4-Flash redrew the budget LLM tier. Here is where Haiku 4.5, GPT-5.4 Nano, and Gemini Flash-Lite now sit.

Qwen3-Next-Thinking: the cheapest reasoning model under $1/M output

OpenAI on AWS Bedrock: what it costs and what actually changed

GitHub Copilot goes metered: what developers will actually pay per 1M tokens

DeepSeek V4 vs GPT-5.5 vs Claude Opus 4.7: what three frontier models actually cost

Gemini Deep Research: what each AI research report actually costs

Kimi K2.6: the fastest model in the top 5, at the lowest price

Claude Code vs OpenAI Codex: which coding agent actually costs less in 2026?

Voice AI APIs in 2026: what Gemini TTS, Voxtral TTS, and OpenAI TTS actually cost per hour

Gemini 3 Flash: $0.50 per million tokens, thinking on by default, and it actually beats Pro on agentic tasks

Tokenmaxxing can inflate your LLM API bill by 10x. On Gemini and GPT-5.4, it's worse.

Claude Opus 4.7: $5 per million tokens - and what that actually means now

Claude Code Routines: what each automated run actually costs

GPT-5.4 computer use: what a real agent task actually costs

DeepSeek V4: $0.30 per million tokens for a 1 trillion parameter model

Chatbot Arena April 2026: Claude leads everything, Grok 4.20 has the cheapest output

OpenAI's new $100 ChatGPT Pro: what you actually get on Codex, and when the API wins anyway

GPT-5.5 "Spud": release date, pricing forecast, and what we actually know right now

LLM API pricing in April 2026: from $0.05 to $125 per million tokens

Meta Muse Spark: no API pricing, no open weights, and one area where it's best in the world

Project Glasswing and Claude Mythos Preview: the AI that found a 27-year-old bug for under $50

Is Claude Code getting worse? The data says something did change on March 8.

Claude Max subscribers using OpenClaw now pay API rates. Here's the math.

Microsoft MAI models: what MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2 actually cost

Gemini 3.1 Pro: $2 input, tied for #1 on benchmarks, and 20% cheaper than GPT-5.4

Qwen3.5-Omni: the pricing, the audio benchmarks, and whether the architecture hype is real

Qwen3.6-Plus: $0.28 per million input tokens, and the benchmark comparison Alibaba chose not to lead with

Claude Haiku 3 retires April 19: it's not just a model ID swap

Gemma 4 is out: $0.14 per million tokens for a 31B model scoring 89% on AIME

Reasoning models in 2026: $0.55 to $20 per million tokens, and when each tier makes sense

Gemini Flex and Priority inference: how Google's new tiers work and what they cost

Google Gemini API billing caps are live: what developers need to know

OpenAI Deep Research API: what it costs, and why o3-deep-research is 5x pricier than o3

OpenAI killed Sora. The math explains why.

ARC-AGI-3: the benchmark no AI can crack, and what running it costs

Gemini 2.0 Flash is deprecated: what migration actually costs you

Kimi K2.5 vs GPT-5.4: the model Cursor built on, and what it actually costs

OpenAI Codex pricing: API costs, container billing, and how it stacks up against Claude Code

How much does Claude Code actually cost per session?

Claude Mythos pricing: what Anthropic's leaked new model will cost developers

Grok 4.20 Beta: $2 per million tokens, 2M context, and the lowest hallucination rate measured so far

Llama 4 Scout vs Maverick: API pricing, self-hosting costs, and which one to use

DeepSeek V3.2 vs GPT-5.4: Is the 30x price gap worth it?

Qwen3.5 Small: the 9B model that beats gpt-oss-120B on four benchmarks

Anthropic drops the 2x long-context surcharge: what Claude now costs at 1M tokens

Xiaomi MiMo-V2-Pro: the trillion-parameter model that fooled everyone into thinking it was DeepSeek

Gemini 3.1 Flash-Lite: $0.25 per million tokens, 1M context, and benchmark scores that beat Claude Haiku

Mistral Small 4: $0.15 per million input tokens for a multimodal MoE model

GPT-5.4 Mini vs Nano: pricing, benchmarks, and which one to use

Claude 5 release date: what Anthropic has actually said