OpenCode says its DeepSeek V4.1 Flash configs match DeepSeek's API, after reports the model feels dumber

OpenCode's founder (thdxr) wrote that OpenCode Go gets occasional reports that DeepSeek V4.1 Flash "feels dumber" for some people. He said he ran every benchmark he could find against the different configurations OpenCode Go serves, and that the results are in the same range.

According to the chart he posted, the configurations score 89.3 to 90.8 overall, against 90.5 for the official DeepSeek API with thinking on. Two low-traffic configurations (about 3% and 9% of traffic) fall to 66.7 and 73.3 on 100k-token context. These are OpenCode's own measurements.

He said the cause in those two was the upstream provider applying a content filter that fired far too often, so affected users got an error rather than a weaker model, and both configurations are being removed. He added that the other issues people report are probably real, asked for specific examples because most reports so far accuse OpenCode of scamming users, and said tokens per second are slower because of limited hardware, which should improve as capacity grows.

The post had about 393 likes on X and 46 replies within its first hour.