Models
-
Same card, new defaults
Google priced Gemini 3.8 Live with the preview it replaces, and moved four defaults under the model string
Two voice models went generally available on 15 September on one rate card with gemini-3.1-flash-live-preview. Swap the string and asynchronous tool calls, always-on proactive audio and video on every turn come with it.
On the preview with tools that assume a pause: declare behavior: BLOCKING on each tool before you change the model string. On the preview at thinking level high: move to Extended Thinking, not 3.8 Live, and rewrite the client to poll interaction_status rather than stop on turnComplete.
-
One request, no reply
Anthropic’s API will now write the summary itself and hand back a signed block that replaces the messages
A beta dated 14 September adds a compaction parameter: send the conversation, get one block and no answer, then send the block first. Leave the old messages in front of it and the next call is an HTTP 400.
If your agent loop compacts by hand, switch to compaction: {type: summarize} on a background request and swap the signed block in; remove the summarised messages or the next call fails with a 400. On Bedrock, Vertex or Foundry, nothing changes yet.
-
Two days on
DeepSeek’s reprieve footnote is still on the page, and the resale markup is still there too
DeepSeek’s pricing page still carries footnote two, read again on 16 September, continuing V4 Pro “with the billing method remaining unchanged”.
If you buy V4 Pro through OpenRouter and can wait, the batch slug costs what DeepSeek charges directly.
-
The reroute that wasn’t
DeepSeek has called off the reroute, and V4 Pro answers its own name after all
A sentence added to a change-log entry dated 10 September says the model continues “in response to user demand”. Its rate card did not move, and this paper printed the cancellation as a certainty in its calendar.
If you moved off deepseek-v4-pro last week, move back. DeepSeek’s own off-peak rate of $0.66 per million input tokens undercuts every OpenRouter route to the same checkpoint, the cheapest of which bills $0.983. If you set a provider order on OpenRouter’s 0813 slug to exclude DeepSeek, that ordering is now routing you around the cheapest seller. Pin the dated slug wherever a host offers one, because the undated name is the one DeepSeek has already once proposed to move.
-
Nothing filed
Neither OpenAI’s nor Anthropic’s developer changelog has an entry after 10 September
Two pages that normally carry the week are empty for it.
Nothing to do. If you are holding a migration for a model announcement, the announcement did not come this week.
-
Three weeks
OpenAI gave gpt-5.4-cyber a shutdown date of 1 October
OpenAI added a deprecation row dated 11 September: gpt-5.4-cyber leaves the API on 1 October and the named replacement is gpt-5.6-cyber. Unlike the Sora row further down the same page, the replacement column is filled.
Grep for gpt-5.4-cyber and move it to gpt-5.6-cyber before 1 October.
-
The name stayed
Calls to DeepSeek V4 Pro reach a different model from Monday, and DeepSeek’s own card shows what it is worse at
At 04:00 UTC on 14 September the name deepseek-v4-pro starts returning V4.1 Flash at under a quarter of the price. Neither DeepSeek’s pricing page nor its change log offers a way to keep the old model.
Search code, configs and gateway routing tables for the string deepseek-v4-pro today: anything that matches changes model at 04:00 UTC on Monday whether or not you deploy. On OpenRouter, give the 0813 slug a provider order that leaves DeepSeek out until its listing says what it serves.
-
In brief
GPT-Live 1 moved from ChatGPT into the API at five cents a minute, with the thinking billed on top
OpenAI made gpt-live-1, ChatGPT’s default voice model since 8 July, generally available in the API on 10 September at $0.05 a minute, billed per second. OpenAI says it holds a full-duplex conversation “while a backend model or agent handles reasoning and tools,” and that backend bills separately at its own rates. The verdict: five cents buys the voice, and the thinking is a second meter.
Before comparing this with a voice stack you run yourself, work out the backend tokens a minute of your conversations burns.
-
Beside the lead
GLM-5.3-Flash went back to list price, and both halves of the bill doubled
Z.ai’s half-price promotion on its cheap tier ended at midnight Beijing time on 9 September.
If GLM-5.3-Flash sits anywhere in your routing table, your cost model for it is now exactly twice wrong. Re-run the comparison against the other cheap tiers before assuming it still wins on price.
-
Beside the lead
Solar Pro 4’s discount ends in three steps, not the one cliff the trackers printed
Upstage’s own card puts the model at $0.09 per million input tokens from 11 September, and at $0.30 only from 10 October.
Anyone who budgeted a tenfold cost cliff for this week has a month of runway at threefold instead. Read the vendor card before you act on any aggregator’s price calendar, including on the models where its number looked right.
-
In brief
IBM relicensed its time-series forecaster off non-commercial, then said it will not maintain it
Granite PatchTST-FM-r2 ships under Apache 2.0 or OpenMDW 1.0, the user’s choice. Its predecessor r1 carries `cc-by-nc-sa-4.0`, which kept it out of products. The card also states IBM “will not be maintaining this code going forward.” The repository was created on 7 August, a month before the announcement. IBM places its CRPS geometric mean of 0.467 second on GIFT-Eval among replicable zero-shot models, behind one entrant it does not name.
If you shelved r1 on licence grounds, re-run that evaluation. Then price the absence of a maintenance commitment into the decision, because the licence is the only thing you are being given.
-
The price moved
Mercury 2.5’s launch discount ended at 07:00 UTC today, and the price is five times what it was yesterday
Inception’s 80 per cent launch discount on Mercury 2.5 Preview, routed through its own provider on OpenRouter, ran through 8 September at 07:00 UTC. Input goes to $0.20 per million tokens from $0.04. Output goes to $0.75 from $0.15, a uniform five times, and Inception’s site lists the same undiscounted rates. The output price is now identical to Mercury 2’s $0.75, so the new model’s premium is on input only, $0.20 against $0.25. What the five times buys over Mercury 2, on Inception’s own description, is tunable reasoning levels and a ten-point intelligence jump that Inception scores itself.
Re-check any budget that priced Mercury 2.5 at the promo rate; the discount was on the Inception route only, so other providers’ prices did not move.
-
The licence
Nvidia’s new diarization model forbids you from publishing what you measure
Nvidia posted Nemotron-3-Diarization-preview on 5 September under an evaluation licence that requires the company’s prior written consent before a user discloses benchmark results.
You can evaluate this, you cannot ship it, and you cannot publish the numbers you get without asking Nvidia. Read the licence before you spend a sprint on the evaluation, not after.
-
Open weights
A diffusion language model with delete and insert tokens nearly doubled its tool-calling score
inclusionAI’s LLaDA2.2-mini reached 47.68 per cent on BFCL v4 function calling, up from 25.05 per cent for LLaDA2.0-mini. It runs 1.4B active parameters against 16B total, over 256 experts. DELETE and INSERT control tokens let it rewrite spans in place rather than decode strictly left to right, and it ships Apache-2.0.
-
Sampling
Gemini stopped sampling video at a fixed frame rate and cut token use by up to 88%
Gemini 3.7 Flash, 3.6 Flash and 3.5 Flash-Lite now sample video adaptively rather than at a fixed frame rate. Google reports token use down as much as 88% against fixed-rate sampling, cost down as much as 66%, and accuracy up as much as 7%.
If you process long video through the Gemini API your bill has already moved. Re-measure before budgeting the next quarter on last quarter’s rate, and do not port the 88% figure into a plan as though it were yours.
-
Open weights
Qwen rebuilt 37,300 runnable terminal environments out of agent logs
Qwen researchers replayed terminal-agent trajectories to rebuild 37,300 runnable environments. Fine-tuning Qwen3.5-27B on the result added 11.9 points on Terminal-Bench 2.1 against the base model.
-
Open weights
A 6B image generator ships with its weights, its code and its recipe
LLaDA-Image pairs a 6B diffusion transformer with a frozen vision-language model. Its weights, code and training recipe ship under CC BY-SA 4.0, against closed pipelines that publish none of the three.
-
Model watch
Google put a vulnerability-scanning Gemini behind a vetting programme
Google shipped Gemini 3.8 Flash Cyber alongside 3.8 Flash. Both are priced for now at $0.75 and $3.75 per million tokens, and both rates double on 1 January 2027.
-
Model watch
DeepSeek’s next vision weights appeared before the announcement
A public Hugging Face repository for DeepSeek-V4-Flash-Vision-Exp carries MIT licensing, FP8 weights, and endpoint-compatible tags. There is not yet enough published evidence to recommend a switch. The news is the sequencing: the usable artifact arrived before the product narrative.
Watch upload times as a leading indicator on launch weeks.