OpenCode has cut the Console price of GLM 5.3 and DeepSeek V4.1 Flash by 60%. The team behind the open-source coding agent announced it on X at 11:14am NZDT on 22 November, and Dax Raad, who builds OpenCode at Anomaly, followed two minutes later with the pitch: OpenCode is now pricing 60% below Fireworks and Baseten “while doing more volume than them”, all on its own GPUs. A token is the small chunk of text a model reads or writes.
What the new prices are
The Console pricing page now shows both models with the old price struck through. GLM 5.3 drops from US$1.40 to $0.56 per million input tokens and from $4.40 to $1.76 per million output tokens, while DeepSeek V4.1 Flash drops from $0.30 to $0.12 on input and from $1.20 to $0.48 on output. Cached reads, the cheaper rate for text a model has already seen, fall by the same 60%.
| Model and token type | Before | Now |
|---|---|---|
| GLM 5.3, input | $1.40 | $0.56 |
| GLM 5.3, cached read | $0.26 | $0.104 |
| GLM 5.3, output | $4.40 | $1.76 |
| DeepSeek V4.1 Flash, input | $0.30 | $0.12 |
| DeepSeek V4.1 Flash, cached read | $0.006 | $0.0024 |
| DeepSeek V4.1 Flash, output | $1.20 | $0.48 |

A million input tokens plus a million output tokens on GLM 5.3 now costs $2.32 instead of $5.80, and on DeepSeek V4.1 Flash $0.60 instead of $1.50. GLM 5.3 Flash, the smaller GLM model, keeps its $0.15 and $0.50 price. The docs page lists end dates for its Gemini discounts and gives none for these 60% rows.
How it compares
Fireworks and Baseten both list the two models at the prices OpenCode just dropped from: $1.40 input and $4.40 output for GLM 5.3, and $0.30 and $1.20 for DeepSeek V4.1 Flash. Raad’s 60% is therefore right against list prices. The gap is smaller on one line, since Baseten lists GLM 5.3 cached reads at $0.14 against OpenCode’s $0.104, and Fireworks’ spot tier, which runs 30% under its standard rate, still sits above OpenCode at $0.98 and $3.08 for GLM 5.3 and $0.21 and $0.84 for DeepSeek V4.1 Flash.
DeepSeek’s own API charges $0.30 and $1.20 at peak and half that, $0.15 and $0.60, off-peak, so OpenCode’s $0.12 and $0.48 undercuts the model maker’s off-peak rate. DeepInfra is running promotions that go lower on some lines: DeepSeek V4.1 Flash at $0.14 input and $0.42 output, or $0.098 and $0.294 for the 1017 build, against OpenCode’s $0.48 output. On GLM 5.3 its promo input price of $0.5625 is level with OpenCode’s $0.56, and its $2.50 output price is above OpenCode’s $1.76. OpenCode isn’t the lowest listed price on every line. I read the Fireworks, Baseten, DeepSeek and DeepInfra pages on 22 November NZ.

What the post says beyond price
Raad says the prices work because everything runs on OpenCode’s own hardware. OpenCode said on 19 November that its own GPUs served 54% of its tokens that day, and Raad wrote on 17 November that it had 1,000 GPUs on short-term contracts. Neither post breaks out which models run on that hardware, so the 54% can’t be split between these two models and the rest.
The volume claim comes with no numbers either. OpenCode’s console page says it serves 600 trillion tokens a month, and Raad’s post gives no figure for Fireworks or Baseten to compare it with.

Discover more from Sulat AI & Tech News
Subscribe to get the latest posts sent to your email.




