Differences

This shows you the differences between two versions of the page.

View differences: Side by SideInline Go

Link to this comparison view

Both sides previous revisionPrevious revision 2026/04/11 16:19 amolith beeg brane here2026/04/11 15:01 kat 2026/04/11 11:21 xenolandscapes [Step 2: Optimize Your Workflow] 2026/04/11 11:15 xenolandscapes [Step 1: Check Your Tools] 2026/04/11 05:29 kat 2026/04/11 01:09 kat 2026/04/10 16:22 amolith add "why model not in harness list" section2026/04/10 01:01 kat [When will Synthetic get model X?] 2026/04/10 00:53 xenolandscapes [When will Synthetic get model X?] 2026/04/10 00:32 kat [When will Synthetic get model X?] minor grammar2026/04/10 00:29 xenolandscapes [When will Synthetic get model X?] 2026/04/10 00:29 xenolandscapes [When will Synthetic get model X?] 2026/04/10 00:27 xenolandscapes createdGo Next revision		Previous revision 2026/04/11 16:19 amolith beeg brane here2026/04/11 15:01 kat 2026/04/11 11:21 xenolandscapes [Step 2: Optimize Your Workflow] 2026/04/11 11:15 xenolandscapes [Step 1: Check Your Tools] 2026/04/11 05:29 kat 2026/04/11 01:09 kat 2026/04/10 16:22 amolith add "why model not in harness list" section2026/04/10 01:01 kat [When will Synthetic get model X?] 2026/04/10 00:53 xenolandscapes [When will Synthetic get model X?] 2026/04/10 00:32 kat [When will Synthetic get model X?] minor grammar2026/04/10 00:29 xenolandscapes [When will Synthetic get model X?] 2026/04/10 00:29 xenolandscapes [When will Synthetic get model X?] 2026/04/10 00:27 xenolandscapes createdGo
faq [2026/04/11 05:29] – kat		faq [2026/04/11 16:19] (current) – beeg brane here amolith
Line 1:		Line 1:
	====== Frequently Asked Questions ======		====== Frequently Asked Questions ======
-	===== When will Synthetic get model X? =====	+	{{namespace>faq}}
-		+
-	Before you ask this, make sure it:	+
-		+
-	- Is [[https://opensource.org/ai/open-weights|open weight]] or [[https://opensource.org/ai/open-source-ai-definition|open source]].	+
-	- Has weights available on [[https://huggingface.co/models|HuggingFace]] (these are not always the same thing: sometimes a model is //planned// to be open weight, but has not been released yet).	+
-	- Has a compatible license that allows Synthetic to actually make money hosting a model (sometimes model weights are published "openly," but only under modified OSS licenses that require royalties over a certain profitability limit, for instance).	+
-	- Has an [[https://developer.nvidia.com/blog/introducing-nvfp4-for-efficient-and-accurate-low-precision-inference|NVFP4]] [[https://ngrok.com/blog/quantization|quantization]] available so that Synthetic can run it on their GPUs at optimal speed (there are some exceptions to this — if a model is sufficiently desired, they may make their own quant).	+
-	- Has solid support for that model or its general architecture in [[https://github.com/sgl-project/sglang/issues|sglang]], the inference engine Synthetic uses to actually //run// the models.	+
-		+
-	Factors that can delay Synthetic getting a model:	+
-		+
-	* If it is unusually large, it may take time for them to acquire or free up GPU space to host it.	+
-		+
-	* If it has a novel or unusual architecture (such as DeepSeek Sparse Attention for GLM 5), it will take time for inference engines like sglang to get reliable support for the model.	+
-		+
-	* If the model has not yet been quantized to NVFP4, Synthetic will have to wait for NVIDIA to do that, or make one themselves, both of which can take some time.	+
-		+
-	Additionally, it is worth keeping in mind that many models from labs known for creating open-weight models may either be closed source—such as Qwen 3.6-Plus—or available only through the lab's API for user testing and feedback (and to give the lab a profitable head start) but not yet released as open weights. This was the case with GLM 5.1 for a few weeks and remains true as of April 9th, 2026, for MiniMax M2.7.	+
-		+
-	===== Why is model X from Synthetic not in $preferred_harness's list? =====	+
-		+
-	Many of these lists are updated by hand by a human, so you might be the first ''$preferred_harness'' user who's noticed one's missing! You can either	+
-		+
-	* Wait until someone else opens a PR to ''$preferred_harness'''s list	+
-	* Find where ''$preferred_harness'' sources its data from and open that PR yourself	+
-	* OpenCode has [[https://github.com/anomalyco/models.dev/|Models.dev]]	+
-	* Crush has [[https://github.com/charmbracelet/catwalk|Catwalk]]	+
-	* Use some kind of provider-specific plugin for ''$preferred_harness'' with a list that updates more frequently	+
-	* Pi has [[https://github.com/ben-vargas/pi-packages/tree/main/packages/pi-synthetic-provider#readme|@benvargas/pi-synthetic-provider]]	+
-	* Maintain your own provider/model list for ''$preferred_harness''	+
-	* [[https://opencode.ai/docs/providers#custom-provider|OpenCode]]	+
-	* [[https://github.com/badlogic/pi-mono/blob/main/packages/coding-agent/docs/models.md|Pi]]	+
-	* [[https://github.com/charmbracelet/crush?tab=readme-ov-file#openai-compatible-apis|Crush]]	+
-		+
-	===== Why am I burning through my credits or requests so quickly? =====	+
-		+
-	==== Step 1: Check Your Tools ====	+
-		+
-	The software you use with Synthetic has a massive impact on your token "burn rate."	+
-		+
-	=== 1. Are you using Claude Code? ===	+
-		+
-	**Recommendation: Use any other agent.** We strongly recommend //against// using Claude Code as a coding harness with Synthetic. Its underlying infrastructure is notoriously inefficient, causing excessive token bloat. Using Claude Code will drain your credits much faster than almost any other option.	+
-		+
-	=== 2. Are you using OpenCode (with oh-my-opencode/oh-my-openagent)? ===	+
-		+
-	While OpenCode is significantly better than Claude Code, it is still not optimized for efficiency. If you are using the oh-my-opencode/openagent (OMO) extension, the problem is significantly worse: OMO launches unnecessary, poorly designed subagent workflows that bloat every single prompt with redundant context, leading to a "death by a thousand tokens" scenario.	+
-		+
-	=== 3. Are you using Zed? ===	+
-		+
-	Zed is a powerful editor, but its real-time "live-edit" feature comes at a cost. Zed utilizes a two-step process for edits:	+
-		+
-	- **Intent to Edit:** It sends a request to define the goal.	+
-	- **Execution:** It runs a //separate// request with the same chat history to generate the code and live-diff.	+
-		+
-	Because of this "intent to edit" system, you are essentially using **2x the input tokens** and **2x the requests** for every single edit. If this workflow is necessary for you, you may need more packs to sustain it.	+
-		+
-	==== Step 2: Optimize Your Workflow ====	+
-		+
-	If your tools aren't known for being wasteful but your usage remains high, follow these steps (roughly in order of recommendation) to reduce token bloat:	+
-		+
-	=== 1. Reduce Frequency ===	+
-		+
-	Reduce the frequency of automated workflow runs. For example, if you use OpenClaw, review your current tasks in OpenClaw's "heartbeat" function and increase the interval between checks.	+
-		+
-	=== 2. Prompt Efficiency ===	+
-		+
-	Refine your input tokens. Use concise system prompts and AGENTS.md files.	+
-		+
-	=== 3. Model Tiering ===	+
-		+
-	Switch to a cheaper model for simpler tasks; Kimi and GLM don't need to be running for every single prompt.	+
-		+
-	=== 4. Limit "Thinking" ===	+
-		+
-	For less complex tasks, reduce the model's budget for thinking. This forces the model to be more direct and prevents it from using tokens on unnecessary internal reasoning.	+
-		+
-	=== 5. Increase Packs ===	+
-		+
-	If your workflow is already lean but you still hit limits, it may be time to upgrade your number of packs to match your professional output.	+