Amp
Amp is an AI coding agent developed by Sourcegraph, designed to assist developers with code writing, debugging, and software engineering tasks. Its core capabilities include autonomous reasoning and multi-step task planning and execution. Amp can understand codebase context and automate complex programming workflows, making it applicable to the daily software development needs of individual developers and engineering teams.
Timeline (last 90 days)
Amp's positioning is not limited to a coding assistant, but also covers requirements gathering before writing code, solution discussion, code review, documentation and communication work.
The Amp team tested an "ultra-fast" mode internally and commented that "people haven't yet seen a model that is both this good and this fast."
Amp is deeply tied to OpenAI, and OpenAI models account for the vast majority of tokens in Amp.
Amp's goal is to make Amp the most cutting-edge agent, and whenever a model gains some capability, to immediately explore how the agent can exploit that capability to the fullest.
The Amp team believes the work most willingly handed directly to an agent is email, specifically reading email, judging importance and drafting replies in the user's tone.
Amp's heavy reliance on a single model vendor (OpenAI) means its capability ceiling and pricing are controlled by others; behind the "no longer counting tokens" selling point are cost subsidies or quota policies, and whether this can be sustained long-term remains unknown.
Amp has not yet officially published complete parameters such as the context window, pricing, and call limits of each of the multiple models.
Amp has entered a partnership with OpenAI, integrating multiple OpenAI models including GPT-6 Astra, Ultrafast, and GPT-Live-1, and supports connecting account systems through Sign in with ChatGPT.
Under multi-model architectures, the billing model is harder to predict: when billing by token, the same input sentence may produce several times the price difference depending on the routing result; if unified account settlement is used, costs are smoothed out, but the product side must bear the losses caused by routing mistakes itself.
Amp has announced support for directly calling Ollama's cloud-hosted models through its BYOK model routing mechanism
4 more timeline events
All Facts (11)
Amp's heavy reliance on a single model vendor (OpenAI) means its capability ceiling and pricing are controlled by others; behind the "no longer counting tokens" selling point are cost subsidies or quota policies, and whether this can be sustained long-term remains unknown.
50%UnverifiedAmp is deeply tied to OpenAI, and OpenAI models account for the vast majority of tokens in Amp.
50%UnverifiedThe Amp team tested an "ultra-fast" mode internally and commented that "people haven't yet seen a model that is both this good and this fast."
50%UnverifiedThe Amp team believes the work most willingly handed directly to an agent is email, specifically reading email, judging importance and drafting replies in the user's tone.
50%UnverifiedAmp's goal is to make Amp the most cutting-edge agent, and whenever a model gains some capability, to immediately explore how the agent can exploit that capability to the fullest.
50%UnverifiedAmp's positioning is not limited to a coding assistant, but also covers requirements gathering before writing code, solution discussion, code review, documentation and communication work.
50%UnverifiedUnder multi-model architectures, the billing model is harder to predict: when billing by token, the same input sentence may produce several times the price difference depending on the routing result; if unified account settlement is used, costs are smoothed out, but the product side must bear the losses caused by routing mistakes itself.
50%UnverifiedUnder Amp's BYOK mode, there are no call count limits and no additional fees are charged
50%UnverifiedAmp supports building remote agents that can be controlled from any location
50%UnverifiedAmp's latest update allows configuring keys from different sources for different tasks
50%UnverifiedAmp has announced support for directly calling Ollama's cloud-hosted models through its BYOK model routing mechanism
50%