Stealth "Pulse Alpha" on VoldemortCode, appears to use GLM tokeniser - GLM 5.5 perhaps?
Summary
OpenCode has launched a mysterious stealth model, Pulse Alpha, offering 1M context, 131K max output, multimodal input, and three reasoning tiers — free for a limited time. Speculation points to GLM 5.5 or the Qwen series under the hood.
View Cached Full Text
Cached at: 11/03/26, 02:20 PM
🚨Breaking! OpenCode just launched the mysterious new model Pulse Alpha! Completely free, 1M context — take off now!
🔹 Context: 1M tokens (million-scale, feed it even the largest repos) 🔹 Max output: 131K tokens 🔹 Multimodal: text + image input (it can see images, but doesn’t natively generate them) 🔹 Reasoning mode: three adjustable levels — low / high / max — plus full tool calling 🔹 Pricing: free for both input and output (limited time, officially for collecting feedback) 🔹 Early reviews: some users say it’s close to DeepSeek V4 Pro 🔹 Speculation: most likely a Qwen-family model (after being jailbroken, it identifies itself as Qwen)
Official positioning: a stealth model, free for now. Data may be retained for improvements. Unavailable in the EU and a few other regions. Just select Pulse Alpha in OpenCode and start using it — still time to jump on it! See the model page for details: https://opencode.ai/v2/docs/console/models/… #OpenCode #PulseAlpha #FreeAI #AICoding #DeepSeek #Qwen #MillionContext
Models | OpenCode
Source: https://opencode.ai/v2/docs/console/models/ OpenCode Console provides a list of models tested and verified by the OpenCode team. Sign in, add credits, and get an API key to use them with OpenCode or another coding agent. Console works like any other provider in OpenCode. It’s completely optional, and you can use any other provider instead. For a subscription with included usage, see OpenCode Go.
How it works
- Sign in to the console, add your billing details and credits, and copy your API key.
- Run
/connectin the TUI, choose the OpenCode pay-as-you-go provider, and paste your API key.
/connect
- Run
/modelsto see the available models and select one.
/models
You are charged per request and can add credits to your account.
Endpoints
You can also access the models directly through the following API endpoints. Free-tier chat models require OpenCode; use paid models or Go from other clients.
Each model is listed with its native API, but every text model also works through the other APIs, which Console translates. See Base URLs for the full set.
These AI SDK packages are for applications calling the endpoints directly.
To select a model in OpenCode, use the format opencode/:
opencode.jsonc
{
"$schema": "https://opencode.ai/config.json",
"model": "opencode/gpt-6.1-sol",
}
Model list
Fetch the full list of available models and their metadata from the models endpoint:
curl https://opencode.ai/inference/v1/models
Jev
Jev is a System One model from TypeSafe AI for fast, structured decisions. Instead of generating text, it evaluates a state against typed questions and returns values and probabilities that your code can use directly. It supports yes/no (noul), multiple-choice (choice), and rubric-based (score) questions.
Use your OpenCode Console API key with the https://opencode.ai/inference/systemone/v1/systemone endpoint.
This example checks whether a support request is urgent:
curl https://opencode.ai/inference/systemone/v1/systemone \
-H "Authorization: Bearer $OPENCODE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "jev-1.13",
"state": "My payments have failed for three days and I am losing sales. Please help now.",
"questions": {
"is_urgent": {
"type": "noul",
"instructions": "Does this request require urgent attention?"
}
}
}'
You can ask multiple questions in one request. Jev evaluates them in parallel and returns each answer under the matching question ID:
curl https://opencode.ai/inference/systemone/v1/systemone \
-H "Authorization: Bearer $OPENCODE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "jev-1.13",
"state": "My order arrived late and the item is damaged. I want a refund.",
"questions": {
"department": {
"type": "choice",
"instructions": "Which team should handle this request?",
"criteria": {
"returns": "Refunds, exchanges, or damaged items",
"shipping": "Delivery delays or lost packages",
"billing": "Charges, invoices, or payment problems"
}
},
"frustration": {
"type": "score",
"instructions": "How frustrated is the customer?",
"criteria": ["Calm", "Frustrated", "Very angry"]
}
}
}'
Use jev-1.13-free instead of jev-1.13 to use the limited-time free model.
See the TypeSafe AI documentation for details about question types and response fields.
Pricing
Console uses pay-as-you-go pricing. Below are the prices per 1M tokens.
Gemini 4 Argon: The 50% discount applies through November 30, 2026.
You may notice low-cost models, such as Haiku, Nano, or Flash, in your usage history. OpenCode uses these models to generate session titles.
Free models
- GLM-5.3-Flash Free is free on OpenCode for a limited time. Its provider follows a zero-retention policy and does not use your data for model training.
- DeepSeek V4.1 Flash Free is free on OpenCode for a limited time. Its provider follows a zero-retention policy and does not use your data for model training.
- MiMo-V2.6-Flash Free is available on OpenCode for a limited time. The team is using this time to collect feedback and improve the model.
- Nemotron 3 Ultra Free is available on OpenCode for a limited time. The team is using this time to collect feedback and improve the model.
- Nemotron 3.5 Lightning Free is available on OpenCode for a limited time. The team is using this time to collect feedback and improve the model.
- Nemotron 3.5 Super VL Free is available on OpenCode for a limited time. The team is using this time to collect feedback and improve the model.
- Big Pickle is a stealth model that’s free on OpenCode for a limited time. The team is using this time to collect feedback and improve the model.
- Pulse Alpha Free is a stealth model that’s free on OpenCode for a limited time. The team is using this time to collect feedback and improve the model. It’s not available in the European Union, Iceland, Liechtenstein, Norway, Cuba, Iran, North Korea, Russia, Ukraine, or Venezuela.
- Beam Free is free on OpenCode for a limited time. Its provider does not use your data for model training and retains it for 30 days for safety reasons. It’s not available in Angola, Belarus, Cambodia, China (including Hong Kong and Macau), Cuba, the Democratic Republic of the Congo, Equatorial Guinea, Iran, Iraq, Laos, Libya, Nigeria, North Korea, Russia, Syria, Turkmenistan, Uzbekistan, Venezuela, or the Crimea, Donetsk, and Luhansk regions of Ukraine.
- Flint 40B Free is free on OpenCode for a limited time. The team is using this time to collect feedback and improve the model.
- Jev 1.13 Free is available on OpenCode for a limited time.
To hide free models in OpenCode, set OPENCODE_DISABLE_FREE_MODELS=1. See free models.
Contact us if you have any questions.
Auto-recharge and budgets
Auto-recharge buys credits when your balance runs low, and budgets cap monthly spend for the workspace and each member.
Deprecated models
Privacy
All our models are hosted in the US and EU. Our providers follow a zero-retention policy and do not use your data for model training, with the following exceptions:
- Big Pickle: During its free period, collected data may be used to improve the model.
- Pulse Alpha Free: During its free period, collected data may be retained and used to improve the model.
- Flint 40B Free: During its free period, collected data may be used to improve the model. Data is retained for 30 days.
- MiMo-V2.6-Flash Free: During its free period, collected data may be used to improve the model.
- Nemotron 3 Ultra Free (NVIDIA free endpoints): Trial use only — do not submit personal or confidential data. Your use is logged for security purposes and to improve NVIDIA products and services. The logged session data for improvement purposes is not linked to your identity or any persistent identifier. For more information about our data processing practices, see our Privacy Policy. By interacting with this endpoint, you consent to our collection, recording, and use of such information and the NVIDIA API Trial Terms of Service.
- Nemotron 3.5 Lightning Free (NVIDIA free endpoints): Trial use only — do not submit personal or confidential data. Your use is logged for security purposes and to improve NVIDIA products and services. The logged session data for improvement purposes is not linked to your identity or any persistent identifier. For more information about our data processing practices, see our Privacy Policy. By interacting with this endpoint, you consent to our collection, recording, and use of such information and the NVIDIA API Trial Terms of Service.
- Nemotron 3.5 Super VL Free (NVIDIA free endpoints): Trial use only — do not submit personal or confidential data. Your use is logged for security purposes and to improve NVIDIA products and services. The logged session data for improvement purposes is not linked to your identity or any persistent identifier. For more information about our data processing practices, see our Privacy Policy. By interacting with this endpoint, you consent to our collection, recording, and use of such information and the NVIDIA API Trial Terms of Service.
- OpenAI APIs: Requests are retained for 30 days in accordance with OpenAI’s Data Policies.
- Anthropic APIs: Requests are retained for 30 days in accordance with Anthropic’s Data Policies.
- Mistral APIs: Prompts and other inputs are not used for training. Data is retained in accordance with Mistral’s Data Retention Policy.
- Beam Free: Prompts and other inputs are not used for training. Data is retained for 30 days for safety reasons.
For subscription models, see Go privacy.
For teams
Share Console models with your team from a workspace. Owners and admins can invite members, choose which models are enabled from the Models tab, set budgets, and connect their own providers.
Background
There are many models available, but only a few work well as coding agents. Providers also vary in configuration, performance, and quality, so a model accessed through a gateway such as OpenRouter may perform differently.
To build this catalog, we:
- Tested a select group of models and talked to their teams about how to best run them.
- Worked with providers to make sure the models were served correctly.
- Benchmarked each model/provider combination to produce a list we feel good recommending.
Console is an AI gateway that gives you access to these models.
Goals
We built Console to:
- Benchmark the best models and providers for coding agents.
- Provide the highest quality options without downgrading performance or routing to cheaper providers.
- Pass along price drops by selling at cost, with markup only to cover processing fees.
- Have no lock-in: use Console with any coding agent, or use another provider with OpenCode.
SuSu_酥酥👅 (@NFT_Chen): 🚨Holy crap! OpenCode pulled out another big move! Operation Cheepseek Phase 4 is officially live!
The Go plan (just $10/month) bumps DeepSeek V4.1 Flash usage all the way up to $100/month!
From $30 → $60 → $80 → now $100 — the allowance keeps climbing.
Still wincing at your API bills? Now you can grab $100 worth of DeepSeek credits for 10 bucks — jump in eyes closed, you can’t lose!
Similar Articles
New stealth model Ox Alpha look like glm or mimo new model
Ox Alpha is a new frontier AI model featuring a 1M token context window and multimodal input capabilities, built for efficient coding and agentic work, with connections to the GLM model developed by Z.ai.
@Khazix0918: https://x.com/Khazix0918/status/2065790596653183156
Zhipu released the GLM 5.2 model, focusing on coding capabilities, open-source and supporting 1M context. Tests show it approaches Claude Opus 4.8 level in large engineering and coding tasks, but lacks multimodal capabilities and is limited by computational power, resulting in slower speed. The article also mentions Anthropic shutting down Fable 5 and Mythos 5 at the request of the U.S. Department of Commerce, highlighting the contrast between open-source and closed AI.
Ox-Alpha Is GLM?
An analysis using prompt injection reveals that the mysterious OX Alpha model on OpenRouter is likely GLM developed by Z.ai, based on system prompt extraction and compression techniques.
@FeitengLi: LLM 玩的好多技术 Povey 在 Zipformer 里都探索过
Z.ai 推出 GLM-5.3-Flash,这是一个具有 1M 代币上下文窗口的多模态 AI 模型,参数规模为 320B-A18B,并以 MIT 许可证发布。
GLM-5.2 is probably the most powerful text-only open weights LLM
Chinese AI lab Z.ai released GLM-5.2, a 753B parameter open weights LLM with a 1M token context window under MIT license, achieving top scores on the Artificial Analysis Intelligence Index and ranking second on the Code Arena WebDev leaderboard.