After the insanity of August where one specific week saw the release of 10 new exciting LLMs, September has been calm in comparison (so far).
Updates in OpenCode Go:
- (New) DeepSeek V4.1 Flash: next big DeepSeek moment is here. Soft replaces DeepSeek V4 Pro. Fast and affordable with vision. $15 usage only expected to double at least
- (New) Muse Spark 1.3 Contributor: the next iteration of Muse Spark, nearly free but Zuck gets all your data and meta geo-bans most of the world
- (Leveled Up) the excellent GLM-5.3-Flash was raised from $15 to $60 credits permanently, 4x the usage quota. This alone restores OpenCode Go’s great bargain status.
| Model | 5HOURS | 1WEEK | 1MONTH | CREDITS |
|---|---|---|---|---|
| DeepSeek V4.1 Flash | 6500 | 16250 | 32500 | $15 |
| Muse Spark 1.3 | 45300 | 113300 | 226600 | $60 |
| GLM-5.3-Flash | 6320 | 15790 | 31580 | $60 |
One major update that OpenCode Go didn’t get yet as far as I know is a stealth Qwen3.8-Max significant upgrade from a single tweet and no real announcement, the new version is named Qwen3.8-Max-0902 and pushes Max’s capabilities to past Kimi K3 and puts it where Fable/Opus5 are. Qwen and OpenRouter already provide the new version. The benchmarks for it aren’t even on Artificial Analysis yet.
This week’s special Go multiplier is the new DeepSeek V4.1 Flash at 4x. Incredible that a model at the level of Opus5 is essentially free for a while.
Forgot to mention that GLM-5.3-Flash was upgraded from $15 to $60 usage, which is amazing since this model is better than GLM-5.2 and can now last all month on medium/high usage.
Updated post