What Shipped This Week: Five New Models, Cleaner Blog URLs, and a Quieter Week Than Usual

Five new models landed in the Writingmate picker this week, and you can test each of them in chat the day it shows up. Alongside that, we cleaned up the blog so old links resolve to one good article instead of several near-copies. This was a quieter week for new product surfaces, and we would rather say so than pad the list.

Below is what you will notice, how to try it in under a minute, and what we worked on that you will not see directly.

Five new models in the picker

This week the model picker gained Gemini 4 Hydrogen, Nex-N2.6, ERNIE-5.5-Flash-Preview, ERNIE-Light-DT-1.0 and LFM3-1.7B. They are a mixed set. Gemini 4 Hydrogen and the two ERNIE models are aimed at long-context research and structured drafting. Nex-N2.6 is aimed at agentic coding and tool use. LFM3-1.7B is a small model, which makes it interesting for quick, cheap tests.

We built the picker around one idea: you should be able to test a new model with your own prompt before you commit to it. Reading a benchmark table tells you little about how a model handles your contracts, your codebase or your product copy.

To try one, open /chat, click the model name at the top of the conversation, and search for the model. Send the same prompt you already use with your current favorite and compare the answers. Each model also has a launch guide on the blog with a suggested test, so you do not have to design one from scratch.

The screenshot shows the selector as it looks today: a searchable list where new models appear alongside the ones you already use. A caveat: some of these are previews. Preview models can change behavior or pricing without much notice, so treat early results as a first look.

A model catalog that refreshes every day

The five models above reached you without anyone hand-editing a list. Our model data is pulled from OpenRouter and refreshed daily, and it ran every day this week. New releases and price changes flow into the picker and the model directory on their own.

We did this because a model is only useful to you if it is there while people are still talking about it. A catalog that updates once a month means you hear about a model on social media and then wait weeks to try it.

To see the result, open the model directory. It lists what is available, with pricing, so you can sort out which models fit your budget before you open a chat.

The directory is the place to check pricing first. If you use the OpenAI-compatible API, the same models are available there under the IDs shown on each model page.

Cleaner blog URLs: 36 posts now redirect

Over the past months our blog accumulated posts that covered the same head topic, for example several "best AI video generator" articles under different slugs. For readers, that means finding two or three articles that say roughly the same thing, with different dates and different prices.

We looked at which post in each group performed best in Search Console, using clicks first and impressions second over the last three months. That post stays. The other 36 now redirect permanently to it, and internal links across the site point at the surviving article.

If you bookmarked an older post, you will land on the stronger version of the same topic. You do not need to do anything. You can browse the full list from the blog.

We also changed how the daily blog picks topics. Before, it only avoided subjects from the last two weeks, so every few weeks it could write the same head topic again under a new name. It now checks the topics we have already covered and skips them. To give it somewhere to go, we added eight new target groups based on free-tool searches that people already run, so upcoming posts should be about things we have not written yet.

One trade-off: consolidating posts means some older titles disappear from the blog index. If you remember a post that is gone, search the blog for its topic and you will find the version we kept.

A new guide, and a link to Subgrow

We published Best AI Tools for Reddit Marketing in 2026. It sorts eight tools by the job each does: mention alerts, finding threads, drafting replies and running your own subreddit. We read the prices on each vendor's own site on November 10 and noted who should skip each tool.

Two of the eight are ours, Subgrow and Writingmate, and the article says so in its second paragraph. Subgrow is a sibling product from the same team, so we also added a followed link to it in the site footer, next to the existing mentioned.to link.

What we worked on that you will not see directly

Two pieces of work this week do not change a screen, but you may want to know about them.

First, we set up a route that sends requests for five open models, DeepSeek V4.1 Flash, GLM 5.3, GLM 5.3 Flash, Qwen3.8 27B and Qwen3.8 2.4T, to a second provider first and falls back to OpenRouter if anything goes wrong before the answer starts. It is off unless we turn it on. If it is active, you are still charged the catalog price shown on the model page, and you should not see a difference in the answer.

Second, we added a typed catalog of 232 events that describe the steps people take from first visit to payment, so we can learn which early actions predict that someone ends up subscribing. Those events go to our analytics tool only and are not sent to ad pixels.

Fixed this week

• Old duplicate blog URLs no longer split traffic across near-identical articles; they redirect to one.
• The daily blog no longer re-writes a head topic it already covers.
• Internal blog links now point at the surviving article rather than a redirected one.
• Retired blog slugs are refused so a deleted post cannot come back by accident.

Thank you for the feedback and bug reports that keep this list honest, including the ones that did not produce a headline this week. The complete history of everything we have shipped lives on the changelog.

— The Writingmate team